Moral Learning Shaun Nichols: Rational Rules: Towards A Theory of

Full download test bank at ebook ebookmass.
com
Rational Rules: Towards a Theory of

Moral Learning Shaun Nichols
CLICK LINK TO DOWLOAD
https://ebookmass.com/product/rational-rules-
towards-a-theory-of-moral-learning-shaun-
nichols/
ebookmass.com
More products digital (pdf, epub, mobi) instant
download maybe you interests ...
The Evolution Of Moral Progress: A Biocultural Theory

Allen Buchanan
https://ebookmass.com/product/the-evolution-of-moral-progress-a-
biocultural-theory-allen-buchanan/
From Deep Learning to Rational Machines [converted]

Cameron J. Buckner
https://ebookmass.com/product/from-deep-learning-to-rational-
machines-converted-cameron-j-buckner/
From Deep Learning to Rational Machines 1st Edition

Cameron J. Buckner
https://ebookmass.com/product/from-deep-learning-to-rational-
machines-1st-edition-cameron-j-buckner/
Einstein's Unfinished Dream: Practical Progress Towards

a Theory of Everything 1st Edition Don Lincoln
https://ebookmass.com/product/einsteins-unfinished-dream-
practical-progress-towards-a-theory-of-everything-1st-edition-
don-lincoln/
Phenomenology Shaun Gallagher
https://ebookmass.com/product/phenomenology-shaun-gallagher/
Classics of Moral and Political Theory 5th Edition,

(Ebook PDF)
https://ebookmass.com/product/classics-of-moral-and-political-
theory-5th-edition-ebook-pdf/
Moral Error Theory 1st Edition Wouter Floris Kalf
https://ebookmass.com/product/moral-error-theory-1st-edition-
wouter-floris-kalf/
A Translation Theory of Knowledge Transfer: Learning

Across Organizational Borders Kjell Arne Røvik
https://ebookmass.com/product/a-translation-theory-of-knowledge-
transfer-learning-across-organizational-borders-kjell-arne-rovik/
The Uses of Delusion: Why It's Not Always Rational to

be Rational Stuart Vyse
https://ebookmass.com/product/the-uses-of-delusion-why-its-not-
always-rational-to-be-rational-stuart-vyse/
OUP CORRECTED AUTOPAGE PROOFS – FINAL, 15/12/2020, SPi
Rational Rules
Copyright © 2021. Oxford University Press USA - OSO. All rights reserved.
Nichols, Shaun. Rational Rules : Towards a Theory of Moral Learning, Oxford University Press USA - OSO,
Rational Rules
Towards a Theory of Moral Learning
SHAUN NICHOLS
1
3
Great Clarendon Street, Oxford, OX2 6DP,
United Kingdom
Oxford University Press is a department of the University of Oxford.
It furthers the University’s objective of excellence in research, scholarship,
and education by publishing worldwide. Oxford is a registered trade mark of
Oxford University Press in the UK and in certain other countries
© Shaun Nichols 2021
The moral rights of the author have been asserted
First Edition published in 2021
Impression: 1
All rights reserved. No part of this publication may be reproduced, stored in
a retrieval system, or transmitted, in any form or by any means, without the
prior permission in writing of Oxford University Press, or as expressly permitted
by law, by licence or under terms agreed with the appropriate reprographics
rights organization. Enquiries concerning reproduction outside the scope of the
above should be sent to the Rights Department, Oxford University Press, at the
address above
You must not circulate this work in any other form
and you must impose this same condition on any acquirer
Published in the United States of America by Oxford University Press
198 Madison Avenue, New York, NY 10016, United States of America
British Library Cataloguing in Publication Data

Data available
Library of Congress Control Number: 2020941874
ISBN 978–0–19–886915–3
DOI: 10.1093/oso/9780198869153.001.0001
Printed and bound by
CPI Group (UK) Ltd, Croydon, CR0 4YY
Links to third party websites are provided by Oxford in good faith and
for information only. Oxford disclaims any responsibility for the materials
contained in any third party website referenced in this work.
For Sarah and Julia

Contents
Preface ix
Acknowledgments xiii
List of Figures xv
I. RATIONALITY AND RULES

1. Rationality and Morality: Setting the Stage 3
2. The Wrong and the Bad: On the Nature of Moral
Representations 25
II. STATISTICAL LEARNING OF

NORM SYSTEMS
3. Scope 49
4. Priors 82
5. Closure 95
6. Status 109
III. PHILOSOPHICAL IMPLICATIONS

7. Moral Empiricism 139
8. Rational Rules and Normative Propriety 164
9. Rationalism, Universalism, and Relativism 192
10. Is It Rational to Be Moral? 211
References 227
Index 245
Preface
My first book in moral psychology, Sentimental Rules, emphasized the role

of emotions in moral judgment. But I never thought emotions exhausted
moral judgment. There are numerous features of moral judgment that are
hard to explain just by appealing to emotions. Why do we tend to think that
it’s wrong to produce a bad consequence, but not wrong (or not as wrong) to
tolerate such a consequence happening? How do we come to think that
some evaluative claims are universally true but others only relatively true?
What kinds of rules can be learned? How do we determine whether some
novel act is permitted or prohibited?
These are questions that arise for moral psychologists and experimental
philosophers. Most work in these areas aims to uncover the processes and
representations that guide judgments. This is the agenda in discussions
about whether people are incompatibilists about free will (e.g., Murray &
Nahmias 2014), whether moral judgment is driven by distorting emotions
(e.g., Greene 2008), and whether judgments about knowledge are sensitive to
irrelevant details (e.g., Swain et al. 2008). Much less attention has been paid
to historical questions about how we ended up with the representations
implicated in philosophically relevant thought. There are different kinds of
answers to these historical questions. One might offer distal answers that
appeal to the more remote history of the concept. For instance, an evolu-
tionary psychologist might argue that some of our concepts are there
because they are adaptations. Or a cultural theorist might argue that some
of our concepts are there because they played an important role in facilitat-
ing social cohesion. On the more proximal end of things, we can attempt to
determine how the concepts might have been acquired by a learner. Those
proximal issues regarding acquisition will be the focus in this book.¹ I will
argue that we can explain many of the features of moral systems in terms of
¹ Of course proximal and distal issues are not unrelated. For an evolutionary psychologist,
the proposal that a concept is an adaptation will typically be accompanied by the expectation
that the characteristic (proximal) development of the concept is not explicable in terms of
domain-general learning mechanisms (see, e.g., Tooby & Cosmides 1992).
x 
rational learning from the evidence. To locate this in contemporary moral

psychology, a bit of background is in order.
In naturalistic moral psychology, sentimentalism is the dominant view
(e.g., Blair 1995; Greene 2008; Haidt 2001; Nichols 2004c; Prinz 2007), and
there is considerable evidence that emotions have numerous influences on
our moral psychology. Emotions seem to impact our judgments about moral
dilemmas (e.g., Bartels & Pizarro 2011; Koenigs et al. 2007). Emotions seem
to influence the resilience of certain moral rules (Nichols 2004c). Emotions
seem to motivate prosocial behavior (Batson 1991). Emotions seem to
motivate punishment for cheaters (Fehr & Gächter 2002). Sentimentalists
have drawn on these results to argue for philosophical conclusions. To take
what is perhaps the most prominent example, the impact of emotion on
certain kinds of moral judgments has been used to challenge the rationality
of those judgments (e.g., Greene 2008; Singer 2005).
I have counted myself among the sentimentalists, but I’ve also argued that
emotional reactions don’t provide a complete explanation of moral judg-
ment. In particular, I’ve argued that rules play an essential role in our moral
psychology (Nichols 2004c). However, I had no account of how we come to
learn these rules. Many moral rules seem to trade on subtle distinctions. For
instance, from a young age, children treat harmful actions as worse than
equally harmful omissions. Children also judge that it’s wrong to harm one
person to save five others from harm. Children are never explicitly taught
the distinctions to which these judgments conform.
The prevailing explanation for how we come to have such subtle distinc-
tions is nativist. Contemporary moral nativists hold that the best explan-
ation for the uniformity and complexity of moral systems is that moral
judgments derive from an innate moral acquisition device (e.g., Harman
1999; Mikhail 2011). Such views hold that the moral systems we have are
partly constrained by human nature. Just as linguistic nativism proposes
constraints on possible human languages, moral nativism implies that there
are constraints on possible human moralities (Dwyer at al. 2010). Although
nativist accounts have been widely criticized (e.g., Nichols 2005; Prinz 2008;
Sterelny 2010), there has been no systematic alternative explanation for how
children acquire such apparently complex moral systems.
My collaborators and I have been developing such an alternative explan-
ation for the acquisition of moral systems. The inspiration comes from an
unlikely source—statistical learning. Recent cognitive science has seen the
ascendance of accounts which draw on statistical learning to explain how we
end up with the representations we have (Perfors et al. 2011; Xu et al. 2012).
 xi
I’ve come to think that statistical learning provides a promising avenue for
answering central questions about how we come to have the moral repre-
sentations we do.
I will argue that a rational learning approach can explain several aspects
of moral systems, including (i) how people learn to draw the act/allow
distinction given limited evidence, (ii) how people come to have a bias in
favor of act-based rules, and (iii) how people use consensus information as
evidence on whether a moral claim is universally true.
The picture that emerges reveals a starkly different side of moral systems
than traditional sentimentalism. The learning processes invoked are, by
standard accounts, rational. This insulates moral judgment from important
charges of irrationality. For instance, if our deontic judgments depend on
rules, and these rules are acquired via rational inference, then we cannot
fault the process by which the judgment is made. This doesn’t insulate the
judgments from every critique. For instance, the rules themselves might be
defective. But that challenge requires a deeper inquiry into the epistemic
credentials of the rules.
The resulting account also contrasts sharply with nativism. The learning
processes that I will draw on are not specific to the moral domain. Indeed,
statistical learning affords the moral psychologist a diverse empiricist tool-
kit. Moreover, the rational learning account suggests that humans are
flexible moral learners, with no innate constraints on the kinds of rules

that humans can learn.
The view that I defend is obviously rationalist in important ways. But that
doesn’t entail a rejection of the significance of emotions for moral judgment.
Indeed, I continue to think that much of the sentimentalist picture is correct.
Emotions play a critical role in amplifying the rules of morality. This
plausibly holds for online decision-making—rules that resonate with strong
emotions will end up having a greater influence in our decision-making. The
emotional amplification of rules also likely explains the cultural resilience of
certain moral rules. To ignore these influences of emotions is to ignore
fundamental aspects of human morality. A persistent commitment of sen-
timentalists down the ages is that without the emotions, we would have
radically different normative systems than we do. I certainly don’t mean to
retreat from that sentimentalist commitment. However, the fact that emo-
tions are critical to our moral systems doesn’t mean that the role of ration-
ality is negligible. On the contrary, I’ll argue, rational learning provides a
much better explanation than emotions for how we acquire normative
systems in all their complexity. The ultimate view, I think, must be some
xii 
form of rational sentimentalism, where rational learning and emotions both

contribute in key ways to our moral judgments. But in this volume, I want to
emphasize the rational side of things. Although moral judgment and deci-
sion might be distorted in many ways, there’s reason to be optimistic that the
fundamental capacity for acquiring moral rules is rational and flexible. The
way we learn rules is plausibly responsive to the evidence in appropriate
ways, and, at least at some developmental stages, supple enough to adjust to
new rules in the face of new evidence.
Acknowledgments
First, I’d like to thank my collaborators on the empirical studies that form
the center of this book: Alisabeth Ayars, Hoi-Yee Chan, Jerry Gaus, Shikhar
Kumar, Theresa Lopez, and Tyler Millhouse. I owe special debts to Theresa,
Alisabeth, and Jerry. Theresa’s dissertation (Lopez 2013) is the first work that
maintained that Bayesian approaches to cognition might provide an alterna-
tive to Chomskyan accounts of moral cognition. If it hadn’t been for Theresa’s
insightful dissertation, I never would have started a project on moral learning.
Alisabeth worked extensively on the project when she was a graduate student
in psychology at Arizona. She had several key experimental ideas; she was also
incisive on the theoretical issues (as evidenced in Ayars 2016). This project
would have been much worse without her contributions. Finally, Jerry was an
ideal collaborator on the empirical work that we did together. More generally,
Jerry has been an intellectually invigorating colleague and friend. It was my
good fortune to be in the same department with him.
Many friends and colleagues have influenced my thinking on these
matters through conversations, discussions in Q&A, and comments on

some of the chapters. In particular, I’d like to thank Mark Alfano, Ritwik
Agrawal, Cristina Bicchieri, Thomas Blanchard, Selmer Bringsjord, Mike
Bruno, Stew Cohen, Juan Comesaña, Fiery Cushman, Justin D’Arms, Colin
Dawson, Caleb Dewey, John Doris, LouAnn Gerken, Josh Greene, Steven
Gross, Heidi Harley, Toby Handfield, Dan Jacobson, Jeanette Kennett, Max
Kleiman-Weiner, Josh Knobe, Max Kramer, Tamar Kushnir, Sydney Levine,
Jonathan Livengood, Don Loeb, Tania Lombrozo, Edouard Machery,
Bertram Malle, Ron Mallon, Eric Mandelbaum, John Mikhail, Adam
Morris, Ryan Muldoon, Scott Partington, Ángel Pinillos, Dave Pizarro,
Jesse Prinz, Hannes Rakoczy, Peter Railton, Sarah Raskoff, Chris
Robertson, Connie Rosati, David Rose, Adina Roskies, Richard Samuels,
Hagop Sarkissian, Sukhvinder Shahi, Dave Shoemaker, David Sobel, Tamler
Sommers, Kim Sterelny, Justin Sytsma, Josh Tenenbaum, John Thrasher,
Hannah Tierney, Mark Timmons, Bas Van Der Vossen, Steve Wall, Jen
Wright, Jonathan Weinberg, David Wong, and Fei Xu.
All of the empirical studies for this project were funded in part by the
U.S. Office of Naval Research under award number 11492159. I’m grateful
xiv 
to Paul Bello, who was the ONR program officer, for supporting the work, as
well as for numerous helpful discussions about it.
Chapter 6 draws substantially on material from Ayars & Nichols (2020),
Rational learners and metaethics, Mind & Language, 35(1), 67–89. I thank
the journal for permission to reprint that material here.
I spent academic year 2017–18 on fellowship at the Center for Human
Values at Princeton. I’m grateful to the Center and to the University of
Arizona for affording me the opportunity to focus on writing the book. In
addition to freeing up time to write, I got excellent feedback from many
people at the Center, including Stephanie Beardman, Mitch Berman, Liz
Harman, Dylan Murray, Drew Schroeder, Amy Sepinwall, Peter Singer,
Michael Smith, Monique Wonderly, and especially Mark van Roojen. Mark
read and commented on much of the book while I was there, and he’s been a
tireless and wonderful correspondent about these issues ever since.
Walter Sinnott-Armstrong arranged to have his research group, Madlab,
read the first draft of the manuscript. This was incredibly helpful. I’m
grateful to all the lab members for taking the time to read and think about
the manuscript. I’d like to single out several people in the group whose
comments led to changes in manuscript: Aaron Ancell, Jana Schaich Borg,
Clara Colombatto, Paul Henne, J. J. Moncus, Sam Murray, Thomas
Nadelhoffer, Gus Skorburg, Rita Svetlova, and Konstantinos Tziafetas.
Mike Tomasello also participated, which was a delight. And of course I’m
especially indebted to Walter, both for organizing the event and for being
characteristically constructive and indefatigable. Dan Kelly also read and
gave terrific comments on the entire manuscript at a later stage. His careful
attention led to numerous improvements in book.
I had the benefit of three excellent referees for OUP, one of whom was
Hanno Sauer (the other two remained anonymous). Thanks to all of them, and
to Peter Momtchiloff for his characteristically excellent stewardship at OUP.
Victor Kumar first encouraged me to write this book. The book turned
out to be a lot more work than I expected, but I still thank Vic for prompting
me to write it, and for excellent comments along the way. Michael Gill and
I have been discussing issues at the intersection of moral philosophy and
cognitive science for twenty years, and his influence and encouragement has
been central to this work. Finally, I’m lucky to have been able to talk with
Rachana Kamtekar about every sticky philosophical problem in the book,
and everything else besides.
List of Figures
1.1 Stimuli for probabilistic inference task 18

3.1 Potential scopes of rules 53
3.2 Hierarchical representation of hypothesis space, based on similarity
judgments 54
3.3 Representing the sizes of hypotheses 57
3.4 Extension of words represented as subset structure 59
3.5 Sample violations of novel rules 62
3.6 Hypothesis space for Principle of Double Effect 67
3.7 Results on intentional/foreseen study 69
3.8 Alternative hypothesis space for Principle of Double Effect 70
3.9 Set of potential patients for a rule 76
3.10 Schematic depiction of display for parochial norms study 78
3.11 Schematic depiction of display for parochial norms study,
20 percent condition 79
4.1 Complete list of violations for overhypothesis study 90

5.1 The rectangle game 99
6.1 Universalist and relativist models for different patterns of consensus 116
6.2 Correlation between perceived consensus and judgments
of universalism 119
6.3 Results on universalism/relativism for abstract cases, by domain 124
6.4 Different relativist models for split consensus 131
7.1 Empiricist model of learning 144
7.2 Hypothesis space for scope of rules 160
9.1 The Stag Hunt 200
9.2 Choosing sides 200
PART I
RATIONALITY AND RULES
1
Rationality and Morality
Setting the Stage
“Moral distinctions are not derived from reason.” Thus does Hume begin his
discussion of morality in the Treatise. Rather, Hume says, moral distinctions
come from the sentiments. Contemporary work in moral psychology has
largely followed Hume in promoting emotions rather than reason as the
basis for moral judgment (e.g., Blair 1995; Greene 2008; Haidt 2001; Nichols
2004c; Prinz 2007; cf. May 2018; Sauer 2017). While I think moral judgment
is tied to emotions in multiple ways, in this book I want to explore the
rational side of moral judgment. I’ll argue that rational processes play a
critical and underappreciated role in how we come to make the moral
judgments we do. In this chapter, I’ll describe the basic phenomena that
I want to illuminate with a rational learning account, and I will explicate the
primary notion of rationality that will be in play.
1. The Phenomena
Don’t lie. Don’t steal. Keep your promises. These injunctions are familiar
and central features of human moral life. They form part of the core
phenomena to be explained by an adequate psychological account of
moral judgment. Why do we make the judgment that it’s wrong to lie or
steal?
In addition to these specific judgments, an adequate moral psychology
must also explain important distinctions that seem to be registered in lay
moral judgment. For example, people tend to think that producing a bad
consequence is worse than allowing the consequence to occur. Much of the
work attempting to tease out an implicit understanding of these distinctions
is done using trolley cases (Foot 1967; Greene et al. 2001; Harman 1999;
Mikhail 2011; Thomson 1976, 1985). For instance, people tend to say that in
the following case, what the agent does is not permissible.
4  
Footbridge: Is it permissible for Frank to push a man off a footbridge and

in front of a moving boxcar in order to cause the man to fall and be hit by the
boxcar, thereby slowing it and saving five people ahead on the tracks?
By contrast, people tend to say that what the agent does (or rather fails to do)
is permissible:
Footbridge-Allow: Is it permissible for Jeff not to pull a lever that would

prevent a man from dropping off a footbridge and in front of a moving
boxcar in order to allow the man to fall and be hit by the boxcar, thereby
slowing it and saving five people ahead on the tracks?
People also tend to say that what the agent does in the following case is
permissible:
Bystander: Is it permissible for Dennis to pull a lever that redirects a

moving boxcar onto a side track in order to save five people ahead on the
main track if, as a side-effect, pulling the lever drops a man off a footbridge
and in front of the boxcar on the side track, where he will be hit? (Cushman
et al. 2006: 1083–4)
These cases have been taken to suggest that people are sensitive to surpris-
ingly subtle distinctions in their normative evaluations.
If people really are sensitive to these distinctions in their moral judg-
ments, these are relatively high-level psychological phenomena. At an even
higher level, we find that people seem to have systematic judgments about
the nature of morality itself. For instance, people tend to think that moral
claims have a different status than conventional claims. This has been
explored extensively with questions like the following:
Authority dependence: If the teacher didn’t have a rule against hitting,

would it be okay to hit other students?
For actions like hitting, people, including pre-school children, tend to say
that it’s wrong to hit even if the teacher doesn’t have a rule. But for actions
like talking during story-time, people are more likely to say that if the
teacher doesn’t have a rule on the matter, it’s okay to talk during story
time (e.g., Nucci 2001; Turiel 1983). More recently, people have explored the
  :    5
extent to which people think moral claims are universally true, using
questions like the following:
Disagreement: If John and Mark make different judgments about whether

it’s okay to rob a bank, does one of them have to be wrong?
For actions like bank robbery and assault, people tend to say that if two
people make different judgments, one of them has to be wrong, but they do
not tend to say this when it comes to aesthetic claims or matters of taste
(Goodwin & Darley 2008; Nichols 2004a; Wright et al. 2013).
These are the phenomena that I want to investigate. Note that much of
our moral lives is not included here. I won’t try to explain our aversion to
suffering in others, our propensity to guilt and shame, or our use of empathy
and perspective taking in moral assessment. Nor will I try to characterize the
ethical abilities enjoyed by non-human animals. The moral capacities that
I’m targeting are, as far as we can tell, uniquely human. How do we arrive at
these sophisticated judgments, distinctions, and meta-evaluations?
2. A Challenge to Common-Sense Morality

Before setting out my positive story, I want to address briefly the prevailing
skepticism about moral judgment. Moral psychologists often cast lay mor-
ality as critically flawed. There is evidence that moral judgment is com-
promised by incidental emotions, misleading heuristics, and confabulation.
Philosophers have used such evidence to develop debunking arguments
according to which key areas of common-sense ethical judgment are epi-
stemically rotten—they are based on epistemically defective processes (see
Sauer 2017 for discussion of debunking arguments). Debunking arguments
have been developed for both common-sense normative ethics and
common-sense metaethics.
2.1 Debunking Normative Ethics
Perhaps the most familiar debunking accounts draw on dual process theories,
according to which there are two broad classes of psychological processes. System
1 processes tend to be fast, effortless, domain specific, inflexible, insensitive to
new information, and generally ill-suited to effective long-term cost–benefit
6  
reasoning. System 2 processes are flexible, domain general, sensitive to new

information, and better suited to long-term cost–benefit analysis, but they are
also slow and effortful.
On Greene’s dual process account of moral judgment, when we are
presented with the option of pushing one innocent person off of a
Footbridge to save five other innocent people, there is competition between
a System 1 emotional process (screaming “don’t!”) and a System 2 process
that calculates the best consequence (saying “5 > 1, dummy”). The proposal
is that cases like Footbridge trigger System 1 emotions that subvert System 2
utilitarian cost–benefit analysis.
A closely related dual process model comes from Jonathan Haidt (2001).
On Haidt’s social intuitionist account, our moral reactions tend to be driven
by System 1 affectively valenced intuitions. System 2 plays a subsidiary
role—it primarily generates post hoc justifications for our affective intu-
itions (2001: 815). One of the key studies that motivates Haidt’s view
suggests that people will hold on to their moral views even when they are
unable to provide a justification for them. For instance, participants were
presented with a vignette in which siblings Julie and Mark have a consensual
and satisfying sexual encounter, using multiple forms of birth control:
Julie and Mark: Julie and Mark are brother and sister. They are traveling
together in France on summer vacation from college. One night they are
staying alone in a cabin near the beach. They decide that it would be
interesting and fun if they tried making love. At the very least it would be
a new experience for each of them. Julie was already taking birth control
pills, but Mark uses a condom too, just to be safe. They both enjoy making
love, but they decide not to do it again. They keep that night as a special
secret, which makes them feel even closer to each other.
What do you think about that? Was it OK for them to make love?
When presented with this vignette, most participants said that it was not
okay for Julie and Mark to make love. When asked to defend their answers,
participants often appealed to the risks of the encounter, but the experi-
menter effectively rebutted the justifications (e.g., by noting the use of
contraceptives). Nonetheless, the participants continued to think that the
act was wrong, even when they couldn’t provide any undefeated justifica-
tions. A typical response was: “I don’t know, I can’t explain it, I just know it’s
wrong” (Haidt 2001: 814). Haidt interprets this pattern as a manifestation of
  :    7
two processes: the moral condemnation is driven by an affective intuition

(rather than reasoning) and the proffered justification comes from post hoc
rationalizing—confabulation.
As Greene and Haidt characterize System 1 processes, the judgments that
issue from those processes are unlikely to be responsive to evidence. Greene
argues that if System 1 is indeed what leads people to judge that it’s wrong to
push in cases like Footbridge, this provides the foundation for an argument
that challenges the rational propriety of non-utilitarian judgment. Greene
suggests that deontological judgments, like “it’s wrong to push the guy in
front of the train,” are defective because they are insensitive to rational
considerations, in sharp contrast with consequentialist evaluations:
[T]he consequentialist weighing of harms and benefits is a weighing

process and not an ‘alarm’ process. The sorts of emotions hypothesized
to be involved here say, ‘Such-and-such matters this much. Factor it in.’
In contrast, the emotions hypothesized to drive deontological judgment are
. . . alarm signals that issue simple commands: ‘Don’t do it!’ or ‘Must do it!’
While such commands can be overridden, they are designed to dominate
the decision rather than merely influence it. (Greene 2008: 64–5)
Greene maintains that since our deontological judgments derive from emo-
tional reactions that are not responsive to rational considerations, we should

ignore them in normative theorizing (2008; see also Singer 2005: 347).¹
Although there is a diverse array of evidence supporting the view that
emotions play a role in judgments about Footbridge (e.g., Amit & Greene
2012; Bartels & Pizarro 2011; Koenigs et al. 2007), emotions cannot provide
a complete explanation for the basic phenomenon of non-utilitarian moral
judgment. Many dilemmas that people rate as generating very little emo-
tional arousal—e.g., those involving lying, stealing, and cheating—elicit
non-utilitarian responses (see, e.g., Dean 2010). Consider, for instance,
cases of promise breaking. People don’t get emotionally worked up by
vignettes that involve promise breaking, but they still make non-utilitarian
judgments about promise breaking. For instance, in one study, participants
were asked whether it was okay for one person to break a promise in order to
prevent two other people from breaking promises; in this case people
maintained that it was wrong for the first person to break a promise even
¹ For direct responses to this argument, see Berker (2009) and Timmons (2008).
8  
though it would minimize promise breaking overall (Lopez et al. 2009: 310).
So emotion doesn’t seem to be required to make non-utilitarian judgments.
Indeed, the asymmetry between Footbridge and Bystander is found even
when the potential human victims are replaced by teacups (Nichols &
Mallon 2006).²
The fact that people make non-utilitarian judgments in the absence of
significant affect indicates that there must be some further explanation for
these responses. This undercuts debunking arguments that depend on the
view that non-utilitarian judgments are primarily produced by arational
emotional reactions. The fact that we find non-utilitarian judgments without
concomitant affect also exposes the need for a different explanation for the
pattern of non-utilitarian judgment that people exhibit.
2.2 Debunking Metaethics
As noted above, people tend to think that at least some moral claims are
universally true, and they treat aesthetic claims as only true relative to an
individual or group (Goodwin & Darley 2008, 2012). Why is this? Why do
people believe of some moral claims that they are universally true?
Philosophers have offered several explanations for the belief in universalism,
and the best-known proposals serve as debunking explanations. In his

influential treatment, Mackie (1977) proposes a number of non-rational
explanations for the belief in universalism. One idea is that motivational
factors, like the desire to punish or compete, play a distorting role in
generating metaethical judgments (see, e.g., Mackie 1977: 43; see also
Fisher et al. 2017; Rose & Nichols 2019). Another of Mackie’s suggestions
is that the belief in universalism derives from the tendency to project our
moral attitudes onto the world. Relatedly, our emotional reactions toward
ethical violations may persuade us that moral wrongs are universally wrong.
The most direct attack on the propriety of metaethical judgments comes
from a study by Daryl Cameron and colleagues (2013). They presented
subjects with brief descriptions of practices in other cultures (e.g.,
“Marriages are arranged by the children’s parents”). In some cases, these
descriptions were presented on a background displaying a disgusting picture
(unrelated to the content of the description); in other cases, the background
² In addition, recent work indicates that Bystander is just as emotionally arousing as

Footbridge (Horne & Powell 2016).
  :    9
was emotionally neutral. Cameron and colleagues found that when the
description was accompanied by a disgusting picture, participants were
more likely to give universalist responses.³ Such an influence is plausibly
epistemically defective. Cameron and colleagues make this clear by drawing
on the distinction between incidental and integral effects of emotions:
Integral emotions may contain information that should appropriately

influence moral judgments: guilt may signal that you have behaved badly
towards others, and anger may signal that others have behaved badly towards
you (Frank 1988). In contrast, incidental emotions are conceptually
unrelated to subsequent judgments, and so are ethically irrelevant
(Doris & Stich 2005). Whereas incidental emotions may influence moral
judgments, they are not appropriately cited as evidence in the justification
of these judgments (719).
If you are more universalist about arranged marriages because you are
seeing a revolting picture of worms, then you’re being swayed by an epi-
stemically defective process.
I focus on the study by Cameron and colleagues because it has a clean
experimental design, and it provides some of the most direct evidence for
the role of an epistemically defective affective process in judgments of
universalism. However, there is a pressing limitation of the study.

Although the results indicate that there is some influence of epistemically
defective processes, the extent of influence is, for debunking purposes,
trivial. The mean difference in universalist judgments produced by inducing
disgust was only 0.1 on a 5-point scale.⁴ Thus, the strongest debunking
conclusion this study can fund is: “To some slight extent, people are not
justified in their belief that a claim is universally true.” Clearly, we cannot
take these results to show that people’s belief in moral universalism is largely
based on a defective process. The results simply don’t explain much of why
people think moral claims are universally true. As a result, they don’t do
much by way of debunking the belief.
³ Cameron and colleagues used a slightly different universalism measure than the standard
disagreement measure (Section 1). They asked participants to evaluate whether an activity
practiced in other cultures (e.g., “Marriages are arranged by the children’s parents”) is wrong
regardless of the culture in which it is practiced.
⁴ More generally, it turns out that the impact of occurrent emotion on moral judgment is
quite weak (e.g., Landy & Goodwin 2015; May 2014, 2018).
10  
I’ve argued that some of the most prominent debunking arguments are
inadequate. Obviously this is a limited selection of the debunking arguments
that have been made. There is a broader lesson here, though. The most
prominent kinds of arguments that purport to debunk lay ethical judgments
appeal to the distorting effects of occurrent emotions. But many of the ethical
judgments that we want to understand do not seem to be explained by
occurrent emotional processes (see also Landy & Goodwin 2015; May 2014,
2018). So I think there is good reason to be skeptical of the attacks on lay moral
judgment. However, skepticism about these accounts hardly constitutes a
positive defense. Even if the extant debunking arguments fail, that doesn’t
mean lay moral judgment is in good repair. The main work of this book is to
promote a detailed positive defense of the rationality of lay moral judgment.
3. Rationality
3.1 The Many Rationalisms
“Rationalism” is used in several strikingly different ways in philosophy. For

much of this book, the notion of rationality in play will be an evidentialist one
on which a person’s belief is rational or justified just in case it is supported by
their evidence. I will argue that several key distinctions in common-sense

morality are acquired through a process of rational inference based on the
evidence that the learner receives. I will set out the evidentialist notion of
rationality in a bit more detail below, but first, I want to chart several different
notions of rationalism which contrast with evidentialism in important ways.
In metaethics, “rationalism” is often used to refer to a view about the
relation between moral requirements and reasons for action. This action-
focused version of rationalism (sometimes called “moral/reasons existence
internalism”) holds that it is a necessary truth that if it is morally right for a
person to Φ then there is a reason for that person to Φ (Smith 1994; van
Roojen 2015).⁵ This view of rationality, unlike a pure evidentialist view, ties
⁵ Michael Smith distinguishes two versions of this rationalist thesis. The conceptual ration-
alist thesis holds that “our concept of a moral requirement is the concept of a reason for action; a
requirement of rationality or reason.” The substantive rationalist thesis holds that this concep-
tual claim bears out in the world. That is, “there are requirements of rationality or reason
corresponding to the various moral requirements” (1994: 64–5).
  :    11
rationality directly to action. Evidentialist approaches to rationality can, of

course, have rationality inform action, since one’s actions will and should be
guided by one’s rationally acquired beliefs. But evidential rationality only
applies to beliefs, not directly to desires or actions.
In cognitive science, “rationalism” is often used to pick out nativist
views in contrast to empiricist views (e.g., Chomsky 2009). These views
are emphatically not rationalist in the evidentialist sense (see, e.g.,
Mikhail 2011: 32). A leading idea of Chomskyan rationalism is precisely
that there are capacities the acquisition of which cannot be explained in
terms of inference over the evidence. For instance, Chomskyans main-
tain that children’s acquisition of grammar cannot be explained in terms
of children drawing apt inferences from the available linguistic data.
As we will see in Chapter 7, moral Chomskyans hold that there is a
moral grammar the acquisition of which can’t be explained in terms of
evidential reasoning. This will be at odds with much of the story that
I develop.
A third notion of rationalism, which prevailed in moral philosophy in the
Early Modern period, is an a priori notion of rationalism. Mathematics
served as the leading example here. The early moral rationalists maintained
that mathematical truths are a priori, and that many of these truths are self-
evident and accessible to us without relying on any kind of experiential
evidence. Similarly, the early moral rationalists (e.g., Clarke, Locke, Balguy)
held there are a priori moral principles that are self-evident, and these self-
evident principles provide the basis for deductive inferences to further moral
claims (see Gill 2007 for discussion). In Chapter 9, I will take up the relation
between this traditional moral rationalism and the more modest moral
rationalism that I’ll promote.
The evidentialist notion of rationality contrasts with all of the above, but
it is the dominant framework in analytic epistemology. According to the
kind of evidentialism that I’ll be using, S’s belief that P is rational or justified
to the extent that S’s belief that P is responsive to her total relevant evidence.
As I want to use the notion, a person’s belief can be responsive to the
evidence even if she lacks conscious access to the reasoning process. Of
course, when subjects do report their reasoning process, if the process they
report is a process that is responsive to the evidence, this gives us good
reason to think that they are in fact making judgments in ways that are—to
some extent—evidentially rational. Still, such conscious access is not
12  
necessary for responding to the evidence.⁶ Algorithms might be responsive

to the evidence even when unconscious.
In addition to its prominence in analytic epistemology, evidentialism also
coheres reasonably well with the notion of rationality that anchors discus-
sions in naturalized epistemology.⁷ Perhaps the best-known statement of
rationality in that literature comes from Ed Stein:
[T]o be rational is to reason in accordance with principles of reasoning that

are based on rules of logic, probability theory and so forth. If the standard
picture of reasoning is right, principles of reasoning that are based on such
rules are normative principles of reasoning, namely they are the principles
we ought to reason in accordance with. (Stein 1996: 4)
On this familiar description, one must reason “in accordance” with the
principles of probability theory to be rational.
What exactly is it to reason “in accordance” with the rules of logic and
probability theory? Much of the literature on reasoning is conspicuously
vague about this (for discussion, see Nichols & Samuels 2017). This much is
obvious, though: reasoning is a process. To clarify the nature of processes,
and the nature of reasoning processes in particular, we can draw on David
Marr’s influential account of levels of analysis in cognitive science.
3.2 Processes
Marr explains levels of analysis in terms of their guiding questions. The first
level is the computational level, and its basic questions are: “What is the goal
of the computation, why is it appropriate, and what is the logic of the
strategy by which it can be carried out?” The second level, representation
and algorithm, asks: “What is the representation for the input and output,
and what is the algorithm for the transformation?” The third level, that of
hardware implementation, won’t occupy us here, but it asks, “How can the
representation and algorithm be realized physically?” (Marr 1982: 25).
⁶ Insofar as evidentialism requires that beliefs are responsive to evidence, the view is typically
taken to be at odds with simple forms of reliabilism (for discussion, see Goldman & Beddor
2016). I won’t try to engage this issue in the book, but reliabilists might maintain that the
inferences that I promote are justified insofar as they are based on reliable processes.
⁷ Research on heuristics and biases has been used to challenge the idea that people are
rational in this evidentialist way (see, e.g., Stich 1990).
  :    13
Marr illustrates different levels of analysis with the example of a lowly

cash register. At the computational level, the process we find in a cash
register is addition, which is a function that takes two numbers as input
and yields one number, in ways specified by an arithmetic theory (22).
Addition is what the cash register does. The computational level also
concerns the question why addition is used for this device, rather than,
say, division or multiplication? Here, Marr says, we can draw on our
intuitions about what is appropriate for the context. In the context of
exchanging groceries for money, it seems most appropriate to sum the
costs of the items purchased (22–3). Intuitions are just one resource for
gleaning the why of psychological processes. Evolutionary psychologists
emphasize adaptationist considerations to explain why a certain process is
appropriate for a context (Cosmides & Tooby 1995: 1202).⁸
Marr’s second level of analysis involves the actual representations and
algorithms of the process (Marr 1982: 23). The computation of addition
can be carried out in different ways. One dimension of flexibility is the
representational system itself. One might use different kinds of symbols to
represent numbers, e.g., binary, Arabic, or hash marks. The other dimen-
sion of flexibility is the algorithms that are deployed. Importantly, the
kinds of algorithms that are appropriate will be constrained by the kinds of
representations in play. With hash marks, one can use concatenation for
addition (|||| + ||| = |||||||), but obviously this would be a disaster for Arabic
numerals (4 + 3 = 43?). Even if the representations are fixed as, say,
Arabic numerals, different algorithms can be used to carry out addition.
One common algorithm for addition mirrors the “carrying” algorithm
people learn in grade school—add the least significant digits, carry if
necessary, move left, and repeat. Another algorithm for addition uses
“partial sums,” separately adding the 1 place-value column, the 10 place-
value column, the 100 place-value column, and then summing these partial
sums. These are different processes at the algorithmic level but not at the
computational level.
⁸ Although Marr doesn’t mention it, sometimes we might identify what the process is, even if
the purpose is unclear or somehow inapt. Imagine that we observe the cash register receive two
inputs—$2 and $3—and it generates $6 as output; then, when another $2 item is entered, the
output is $12. Eventually it becomes clear that what the machine is doing is multiplication. Why
it’s doing multiplication is not because that’s the right process in this context (based on our
intuitions or evolution). Perhaps the machine was hacked or perhaps there’s a short circuit that
remapped + to *.
14  
3.3 Rational Processes
We can now recruit Marr’s distinctions to impose greater specificity in

characterizing rational processes.⁹ On one construal, for a system to be “in
accordance” with the rules of logic and probability theory is for the process
to execute algorithms that encode the rules of logic and probability theory.
So, for instance, if modus tollens is a proper rule of logic, and a given process
deploys a modus tollens algorithm to transition from ((P ! Q) and ~Q) to
~P, then that process is rational on the algorithmic construal of accordance.
In that case, we can say the transition from input to output is algorithmically
rational.¹⁰
Another way to understand “accordance” with the rules of logic and
probability theory is in terms of the computational-level description. On
that approach, a process accords with logic and probability theory when
what the process computes—the input–output profile of the process—
corresponds to the function of the relevant logico-probabilistic rules. In
that case, we can say that the process is computationally rational. As we
saw above, the computational-level description of a process is neutral about
the actual algorithms involved in the transition from input to output.¹¹
Often in cognitive science, what we really want to capture is the algorith-
mic level process. But it’s also the case that often we settle for less. We can
start by trying to show that a process is rational at the computational level,

with the hope that this can eventually be filled out with an algorithmic
⁹ This discussion is based on joint work with Richard Samuels (Nichols & Samuels 2017).
¹⁰ Richard Samuels and I distinguish strong and weak versions of the thesis that some
cognitive process is in algorithmic accordance with logic and probability theory (Nichols &
Samuels 2017). Strong algorithmic accordance requires that there is an isomorphism between
analytic probability theory and the inferential process being evaluated. The idea is that an
algorithm is only rational if it proceeds as prescribed by the mathematics of probability theory.
Samuels and I suggest that this looks to be an excessively demanding way to characterize
rational psychological processes. Instead, we argue for weak algorithmic accordance, which
allows that a process can be rational when the algorithm implements a good Bayesian approxi-
mation method (2017: 24–5). What makes for a good approximation method might vary by
context. Although I won’t discuss weak algorithmic accordance further in this book, the notion
of weak algorithmic accordance makes it rather easier for a psychological process to count as
rational. By contrast, strong algorithmic rationality imposes severe demands on the requirement
for rationality (2017: 22).
¹¹ Not all processes characterized at the computational level need conform to logic and
probability theory. For instance, part of the visual system of the housefly is described at the
computational-level as having the goal of landing (Marr 1982: 32–3); but it needn’t be the case
that the transition from perceptual stimulus to landing behavior in the housefly conforms to the
laws of probability.
  :    15
analysis (see, e.g., Xu & Tenenbaum 2007b: 270).¹² Making a case that moral
learning is characterized by a process that is rational at the computational
level is a step toward an algorithmic analysis, and often the most we can
hope for at this point is a computational analysis.¹³ However, for some of
the inferences that I will promote in this book, I will suggest that we do have
the beginnings of an algorithmic account, reflected, for instance, in the
explanations offered by the experimental subjects.
4. Statistical Learning
4.1 Statistical Learning
Some kinds of statistical learning involve extremely sophisticated techniques

which require enormous computational resources, but other kinds of stat-
istical learning are humble and familiar forms of inference. Imagine you’re
on a road trip with a friend and you have been sleeping while he drives. You
wake up wondering what state you’re in. You notice that most of the license
plates are Kansas plates. You can use this information to conclude that you are in
Kansas. This is a simple form of statistical learning. You are consulting samples
of license plates (the ones you see), and using a principle on which samples
reflect populations (in this case, the population of license plates). This, together
with the belief that Kansas is the only state with a preponderance of Kansas
plates warrants your new belief that you are in Kansas.
Early work on statistical reasoning in adults indicated that people are
generally bad at statistical inference (e.g., Kahneman & Tversky 1973). For
instance, people seem to neglect prior probabilities when making judgments
about likely outcomes. In a striking experiment, Kahneman and Tversky
(1973) presented one group of participants with the following scenario:
A panel of psychologists have interviewed and administered personality

tests to 30 engineers and 70 lawyers, all successful in their respective fields.
¹² Griffiths and colleagues (2015) suggest that the transition from the computational to the
algorithmic level can be facilitated by an intermediate level which adverts to resource con-
straints.
¹³ Note that we don’t need to qualify that a Marrian rational process is merely pro tanto
rational. This is because computational and algorithmic rationality are defined narrowly in
terms of the function of the process. For example, the algorithm for addition is defined in terms
of a restricted class of inputs and outputs and the dedicated transitions between them.
16  
On the basis of this information, thumbnail descriptions of the 30 engin-

eers and 70 lawyers have been written. You will find on your forms five
descriptions, chosen at random from the 100 available descriptions. For
each description, please indicate your probability that the person described
is an engineer, on a scale from 0 to 100. (241)
Another group of participants got the same scenario, but with the base rates
reversed—in this condition there were said to be 30 lawyers and 70 engin-
eers. Participants were then given the five descriptions (allegedly chosen at
random) mentioned in the instructions. One of the descriptions fits with a
stereotype of engineers:
Jack is a 45-year-old man. He is married and has four children. He is

generally conservative, careful and ambitious. He shows no interest in
political and social issues and spends most of his free time on his many
hobbies which include home carpentry, sailing, and mathematical puzzles.
(241)
With this description in hand, subjects are supposed to indicate how likely it
is (from 0 to 100) that Jack is an engineer. Kahneman and Tversky found
that participants in both conditions gave the same, high, probability esti-
mates that Jack is an engineer. The fact that there were 70 engineers in one
condition and 30 in the other had no discernible effect on subjects’ responses
(1973: 241). Another description was designed to be completely neutral
between the lawyer and engineer stereotype:
Dick is a 30-year-old man. He is married with no children. A man of high

ability and high motivation, he promises to be quite successful in his field.
He is well liked by his colleagues. (142)
When given the description that was neutral with respect to the stereotypes,
obviously participants should go with the base rates; instead, in both con-
ditions, they tended to say that there was a 50 percent chance that the person
was an engineer (1973: 242).
This study is representative of the broad pattern of results in the
Heuristics & Biases tradition, which has exposed numerous ways in which
people make mistakes in statistical reasoning. In the wake of this pessimistic
line of research, a new wave of cognitive psychology celebrates people’s basic
abilities in statistical inference. A wide range of cognitive phenomena have been
  :    17
promoted as instances of rational inference, including categorization (Kemp

et al. 2007; Smith et al. 2002), word learning (Xu & Tenenbaum 2007b), and
parsing (Gibson et al. 2013). In addition, since the late 2000s, work in devel-
opmental and cognitive psychology suggests that children actually have an
early facility with statistical reasoning. I’ll present two sets of findings from this
emerging research.
It is normatively appropriate to draw inferences from samples to popu-
lations when samples are randomly drawn from that population, but typic-
ally not otherwise. To see whether children appreciated this aspect of
statistical inference, Xu and Garcia (2008) had infants watch as an experi-
menter reach into an opaque box (without looking in the box) and pull out
four red ping-pong balls and one white one. In that case, it’s statistically
appropriate to infer that the box has mostly red balls. In keeping with this,
when infants were then shown the contents of the box, they looked longer
when the box contained mostly white balls than when the box contained
mostly red balls. Xu and Denison (2009) then investigated whether the
nature of the sampling made a difference. At the beginning of the task, the
experimenter showed a preference for selecting red balls over white ones.
Then, the experimenter selected balls from an opaque container as in the
experiment reported above. But in this study, for one condition the experi-
menter was blindfolded while in the other she had visual access to the
contents of the box. Xu and Denison found that babies were more inclined
to expect the population of balls to resemble the sample in the blindfolded
condition as compared to when the experimenter could see the balls she was
choosing. It seems that the babies were sensitive to whether or not the
sampling was random. Building on these findings, Kushnir and colleagues
found that children use sampling considerations to draw inferences about
preferences. When a puppet took five toy frogs from a population with few
frogs, the children tended to think the puppet preferred toy frogs; but
children tended not to make this inference when a puppet took five toy
frogs from a population that consisted entirely of frogs (Kushnir et al. 2010).
For a second example, consider another feature of good probabilistic
reasoning: If you have priors (e.g., you know the percentage of the popula-
tion afflicted with a certain disease) then you should use those priors (e.g., in
making inferences from a person’s symptoms to whether they have the
disease); furthermore, if you get new information, you should update the
priors. Girotto and Gonzalez (2008) explored such reasoning in children
using a task with chips of different shapes and colors. Figure 1.1 depicts a
18  
Figure 1.1 Stimuli for

probabilistic inference task
(adapted from Girotto &
Gonzalez 2008: 328)
stylized version of the materials—four black circles, 1 black square and 3

white squares.
In the “prior” task, kids were shown just the square chips and told “I’m
going to put all the chips in this bag . . . . I will shake the bag and I will take a
chip from it without looking.” Then they were told that if the chip was black
they would give a chocolate to Mr. Black (a puppet) and if the chip was white
it would go to Mr. White (another puppet). Then they were asked which
puppet they choose to be. Of the 4th grade school children, 91 percent
correctly chose the puppet most likely to win the chocolate (332). Girotto
and Gonzalez explored whether the child can also update in this kind of task.
The child was shown all eight chips and asked which is more likely to win
(with black advantage 5:3). Children tend to say correctly that black is more
likely to win. Then all eight chips are put in the bag and the experimenter
reaches in. He says, “Ah, listen. I’m touching the chip that I have drawn and
now I know something that might help you to win the game. I’m touching
the chip that I have in my hand and I feel that it is [a square]” (331). The kids
are allowed to revise their judgment, and they tend to answer correctly that
now it’s more likely that white will win (334). Note that children succeed in
these tasks with no training—they produce the correct response immedi-
ately. Subsequent work found that children and adults in two pre-literate
Mayan groups also succeed in these tasks (Fontanari et al. 2014).
These are just two examples of ways in which even young children seem
to make statistically appropriate inferences. We will see several further
examples in the course of the book. The lesson of this work is that despite
the foibles that have been revealed by the Heuristics and Biases tradition,
people, including very young children, possess a substantial competence at
probabilistic reasoning.
4.2 Statistical Learning and Rationality
The dominant paradigm for explaining these results is Bayesian learning

(see, e.g., Perfors et al. 2011). The advocates of this view stress the rational
  :    19
nature of Bayesian inference. For example, Amy Perfors and colleagues

write:
Bayesian probability theory is not simply a set of ad hoc rules useful for
manipulating and evaluating statistical information: it is also the set of
unique, consistent rules for conducting plausible inference . . . . Just as
formal logic describes a deductively correct way of thinking, Bayesian
probability theory describes an inductively correct way of thinking.
(Perfors et al. 2011: 313)
For many experiments, it’s not obvious that the data support the view that
children are engaged in a form of Bayesian updating (see, e.g., Nichols &
Samuels 2017). But there is little doubt that the inferences in the tasks
reviewed in Section 4.1 are plausible candidates for meeting familiar notions
of evidential rationality. Critically, in the above tasks, the child makes
inferences that are appropriate given the evidence to the agent. For instance,
in the ping-pong ball task, the infant is right to infer from a random sample
of mostly red balls that the population is mostly red. All of the evidence
she has supports this conclusion.¹⁴ In addition, these tasks take exactly
zero training. The normatively appropriate pattern appears on the first
(and only) trial. Much prominent work in Bayesian psychology claims
only to be giving an analysis of people’s judgments at the computational

level (e.g., Xu & Tenenbaum 2007b: 270). But at least in some cases, an
algorithmic analysis is also plausible, and one can get evidence for this
from people’s reports of their reasoning process (see, e.g., De Neys
& Glumicic 2008; Ericsson & Simon 1984). It’s likely that for tasks like
the chips task from Girotto and Gonzalez, adults would be quite capable of
articulating the reasoning process they actually go through, which might
well provide evidence that their reasoning process is algorithmically
rational.
¹⁴ One potential worry about the rationality of everyday inferences from samples is that the
samples might be unrepresentative. It is plausible that when people have evidence that the
sample is unrepresentative, if they ignore this in their statistical inferences, their inferences are
rationally compromised. However, when a person has no evidence that a sample is unrepre-
sentative, it seems uncharitable to declare their inferences from the sample to be rationally
corrupt. That is, when there is no evidence that a sample is unrepresentative, it’s reasonable for a
learner to make inferences as if it is representative. Indeed, if we were so cautious as to withhold
inferences on the bare possibility that a sample is unrepresentative, we would rarely make
inferences. To suggest such inferential caution borders on recommending skepticism.
20  
4.3 Statistical Learning and Empiricism
I’ll be promoting a kind of empiricist account of moral learning, in terms of

statistical inference. I’ll discuss this at some length in Chapter 7, but for now
I want to provide a bit of the intellectual background that informs the
discussion. In contemporary cognitive science, the debate between empiri-
cists and nativists is all about acquisition (see, e.g., Cowie 2008; Laurence &
Margolis 2001). Take some capacity like the knowledge of grammar. How is
that knowledge acquired? Empiricists about language acquisition typically
maintain that grammatical knowledge is acquired from general purpose
learning mechanisms (e.g., statistical learning) operating over the available
evidence.¹⁵ Nativists about language acquisition maintain instead that there
is some domain-specific mechanism (e.g., a specialized mechanism for
acquiring grammar) that plays an essential role in the acquisition of lan-
guage. In the case of grammatical knowledge, debate rages on (e.g., Perfors
et al. 2011; Yang et al. 2017). But it’s critical to appreciate that there is some
consensus that for certain capacities, an empiricist account is most plausible
while for other capacities, a nativist account is most plausible.
On the empiricist end, research shows that infants can use statistical
evidence to segment sequences of sounds into words. The speech stream is
largely continuous, as is apparent when you hear a foreign language as
spoken by native speakers. How can a continuous stream of sounds be

broken up into the relevant units? In theory, one way that this might be
done is by detecting “transitional probabilities”: how likely it is for one
sound (e.g., a syllable) to follow another. In general, the transitional prob-
abilities between words will be lower than the transitional probabilities
within words. Take a sequence like this:
happy robin
As an English speaker, you will have heard “PEE” following “HAP” more
frequently than ROB following PEE. This is because “HAPPEE” is a word in
English but “PEEROB” isn’t. This sort of frequency information is ubiqui-
tous in speech. And it could, in principle, be used to help segment a stream
into words. When the transitional probability between one sound and the
next is very low, this is evidence that there is a word boundary between the
¹⁵ Note that empiricists allow that these general-purpose learning mechanisms themselves
might be innate; after all, we are much better at learning than rocks, trees, and dust mites.
  :    21
sounds. In a groundbreaking study, Jenny Saffran and colleagues (1996)

used an artificial language experiment to see whether babies could use
transitional probabilities to segment a stream. They created four nonsense
“words”:
pabiku tibudo golatu daropi
These artificial words were strung together into a single sound stream,
varying the order between the words (the three orders are depicted on
separate lines below, but they are seamlessly strung together in the audio):
pabikutibudogolatudaropi
golatutibudodaropipabiku
daropigolatupabikutibudo
By varying the order of the words, the transitional probabilities are varied
too. Transitional probabilities between syllabus pairs within a word (e.g.,
bi-ku) were higher than between words (e.g., pi-go) (p = 1.0 vs. p = 0.33).
After hearing two continuous minutes of this sound stream, infants were
played either a word (e.g., pabiku) or part word (e.g., pigola). Infants listened
longer (i.e., showed more interest) when hearing the part word, which
indicates that they were tracking the transitional probabilities.¹⁶ This ability
to use statistical learning to segment sequences isn’t specific to the linguistic
domain. It extends to segmenting non-linguistic tones (Saffran et al. 1999)
and even to the visual domain (Kirkham et al. 2002). Perhaps humans have
additional ways to segment words, but at a minimum, there is a proven
empiricist account of one way that we can segment streams of continuous
information into parts using statistical learning.
Nativists can claim victories too, however. Birdsong provides a compel-
ling case. For many songbirds, like the song sparrow and the swamp
sparrow, the song they sing is species specific. It’s not that the bird is born
with the exact song it will produce as an adult, but birds are born with a
“template song” which has important elements of what will emerge as the
adult song. One line of evidence for this comes from studies in which birds
are reared in isolation from other birds. When the song sparrow is raised in
¹⁶ Listening time was measured by how long babies looked towards the source of the sound,
which was either on the left or the right side of the room.
22  
isolation, it will produce a song that shares elements of the normal adult
song sparrow song; the same is true for the swamp sparrow. And critically,
the song produced by the isolate song sparrow differs from the song pro-
duced by the isolate swamp sparrow. This provides a nice illustration of a
nativist capacity. It’s not that experience plays no role whatsoever—the
specific song that the bird produces does depend on the experience. But
there is also an innate contribution that is revealed by the song produced by
isolate birds. The template gives the bird a head start in acquiring the
appropriate song (see, e.g., Marler 2004).
The examples of birdsong and segmentation of acoustic strings show that
it’s misguided to think that there is a general answer to the nativist/empiri-
cist debate. There are numerous nativist/empiricist debates, and it’s import-
ant to evaluate the disputes on a case by case basis. The cases I offer in Part II
of this book are all empiricist learning stories, based on principles of
statistical inference. Importantly, however, the arguments in Part II make
no claim to a thoroughgoing empiricism. The work starts with learners who
already have facility with concepts like agent, intention, and cause. It also
starts with the presumption that learners have the capacity for acquiring
rules. I argue that, given those resources and the evidence available to
children, their inferences are rational. This is all consistent with the nativist
claim that the acquisition of the concept of agent (for example) depends on
innate domain-specific contributions and that there is an innate capacity to

learn rules (Nichols 2006: 355–8).
4.4 A Schema for Statistical Learning Accounts
Let’s now turn to the characteristics of such learning accounts. Suppose we

want to argue that some concept (for example) was acquired via statistical
inference over the available evidence. Several things are needed.
(1) One needs a description of the concept (belief, distinction, etc.) the
acquisition of which is to be explained. We can call the target concept
the acquirendum (Pullum & Scholz 2002). Part of the work here is to
argue that we do in fact have the concept or distinction or belief that
is proposed as the acquirendum (A).
(2) Insofar as statistical learning is a form of hypothesis selection, one
needs to specify the set of hypotheses (S) that the learner considers in
  :    23
acquiring A. This set of hypotheses will presumably include A as well

as competing hypotheses.
(3) One will also need an empirical assessment of the evidence (E) that is
available to the learner. For this, one might consult, inter alia, corpus
evidence on child-directed speech.
(4) The relevant statistical principle(s) (P) need to be articulated. These
principles should make it appropriate for a learner with the evidence
E and the set of hypotheses S to infer A.
(5) Finally, a complete theory of acquisition would tie this all together by
showing that the learner in fact does use the postulated statistical
principle P and the evidence E to select A among hypotheses S.
Few empiricist theories of acquisition manage to provide convincing evi-

dence for all of these components. The last item in particular is well beyond
what most learning theorists hope to achieve. For instance, (5) would
require extremely fine-grained longitudinal analyses of the evidence avail-
able to individual children and their use of that evidence in learning. In place
of this daunting demand, a learning theorist might aim for a weaker goal—to
show that learners are capable of using the relevant kind of evidence to make
the inferences that would be appropriate given the postulated statistical
principles. That is, instead of trying to capture the learner’s actual acquisi-
tion of the concept, one might settle for something a good deal weaker:
(5*) Show that when the learner is given evidence like E, she makes
inferences that would be appropriate if she were deploying the postulated
statistical principles P.
This requirement is of course weaker than (5) in that it doesn’t try to show
the actual transition that occurs when a child acquires the concept. Rather,
the goal is to show that learners are appropriately sensitive to the evidence.¹⁷
In addition, (5*) is intended to make a weak claim about the precision and
accuracy of the probabilistic representations. I don’t argue (I don’t even
believe) that people make precisely accurate probabilistic inferences from
the evidence. Rather, the goal of this book is to argue that, for a range of
important elements in our moral psychology, when people learn those
¹⁷ The term “sensitive” has a technical meaning in analytic epistemology (e.g., Nozick 1981),
but I intend the ordinary notion on which being sensitive roughly means responding appro-
priately under different conditions.
24  
elements, they make inferences that are roughly the kinds of inferences that
they should make, given the evidence.¹⁸
Our next task is to determine the kinds of representations that are
implicated in moral judgment. That will be crucial to characterizing the
acquirenda, and it will be the focus of the next chapter.
¹⁸ This modest defense of human rationality is reflected in some of Tania Lombrozo’s work.
For instance, in her lovely work on simplicity, she shows that people are responsive to evidence
in a Bayesian fashion but they overweight the importance of simplicity (2007).
Another random document with
no related content on Scribd:
salir desta miseria? pues dime,
Gallo, en qué tengo de
convertirme despues de que deje
de ser Micillo?
Gallo.—Eso yo no lo sé porque
está por venir; mas volviendo á mi
propósito, como al prencipio de mi
ser yo fuese Euforbio y pelease
ante los muros de Troya matóme
Menelao y dende á poco tiempo
vine á ser Pitágoras; por cierto
vine á vevir sin casa ni techo
donde pudiese posar hasta que
Menesarca me la edificó.
Micillo.—Ruégote que me digas,
¿hacias vida sin comer ni beber?
Gallo.—Por cierto no usaba de
más de lo que al cuerpo le podia
bastar.
Micillo.—Pues primero te ruego
me digas lo que en Troya pasó y
lo que viste siendo tú Euforbio,
por ver si Homero dijo verdad.
Gallo.—¿Cómo lo podia él
saber, pues no lo vio? que cuando
aquello pasaba era él camello en
las Indias; una cosa quiero que
sepas de mí; que ni Ayax Telamon
fue tan esforzado como lo pinta
Homero ni Helena tan hermosa
porque ya muy vieja era, casi
tanto como Hécuba, porque esta
fue mucho antes robada de Teseo
en Anfione; ni tampoco fue tan
elegante Archiles (sic) ni tan
astuto Ulises, que en la verdad
fabula es y muy lejos de la
verdad, como suele acaescer que
las cosas escritas en historias y
contadas en lejos (sic) tierras
sean muy mayores en la fama y
mas elegantes de lo que es
verdad. Esto te baste de Euforbio
y de las cosas de Troya.
CAPITULO VII
Que siendo Pitagoras lo que le

acaesció.
Gallo.—Vengo á contar lo que

siendo Pitágoras me acaesció y
porque cumple que digamos la
verdad, yo fue en suma un sofista
y no nescio, muy poco ejercitado
en las buenas disciplinas, e
acordé de me ir en Egito por
disputar con los filosofos en sus
altas ciencias, con los cuales
deprendí los libros de la diosa
Ceres la qual fue inventadora de
la astrología y primera dadora de
leyes, y despues volvime en Italia,
donde comenze á enseñar á los
latinos aquello que deprendí de
los griegos y de tal suerte doctriné
que me adoraban por Dios.
Micillo.—Ya yo he oido eso y
cómo de los italos fueste creido;
mas dime agora la verdad; ¿qué
fue la causa que te movió que
constituyeses ley que no
comiesen carne ni habas ningun
hombre?
Gallo.—Aunque tengo
vergüenza de lo decir, oirlo has,
con tal condicion que lo calles; yo
te hago saber que no fue causa
alguna ni cosa notable ni de gran
majestad; mas miré que si yo
enseñaba cosas comunes y viejas
al vulgo no serian de estimar; por
tanto acordé de inventar cosa
nueva y peregrina á los mortales
porque más conmoviese á todos
con la novedad de las cosas de
admiracion; ansi yo procuré de
inventar cosa que denotase algo,
mas que fuese á todos incónita su
interpretacion y en conjeturas
hiciese andar á todos atónitos sin
saber qué quería decir, como
suele acaescer de los oráculos y
profecías muy oscuras.
Micillo.—Dime agora, despues
de que dejaste de ser Pitagoras,
¿en quién fuistes transformado y
qué cuerpo tomaste?
CAPITULO VIII
Como siendo Pitágoras fue

transformado en Dionisio rey
de Sicilia y lo que por mal
gobernar se sucede.
Gallo.—Despues sucedi en el
cuerpo de Dionisio rey de Secilia.
Micillo.—¿Fueste tú aquel que
tuvo por nombre Dionisio el
tirano?
Gallo.—No ese, mas su hijo el
mayor.
Micillo.—Pues di la verdad, que
tambien fueste algo cruel y aun si
digo mas no mintiré; tú ¿no
mataste á tus hermanos y
parientes poco á poco porque
temías que te habian de privar del
reino? bien sé que sino te
llamaron el tirano fué porque en el
nombre difirieses de tu padre;
basta que te llamaron siracusano
por las crueldades que heciste en
los siracusanos; dime la verdad,
que ya no tienes que perder.
Gallo.—No te negaré algo de lo
que pasó desde mi niñez, porque
veas el mal reinar á qué estado
me vino á traer. Yo fue el mayor
entre los hijos de mi padre y como
el reinado se adquirió por tirania
no sucedimos los hijos herederos,
sino trabajabamos ganar la gente
del pueblo que nos habia de
favorescer, y ansi yo procuré
quanto á lo primero haber á pesar
de mis hermanos los tesoros de
mi padre, con los cuales como
liberal distribuí por los soldados y
gente de armas, que habia mucho
tiempo que mi padre los tenia por
pagar, y despues por atraer el
pueblo á mi favor solté tres mil
varones que mi padre tenia en la
carcer muy miserablemente
atados porque no le querian
acudir con sus rentas y haciendas
para aumentar sus tesoros y
solteles el tributo por tres años á
ellos y á todo el pueblo. Mas
despues que fue elegido de los
ciudadanos y comarcanos, ¡oh
Micillo! vergüenza tengo de te lo
decir.
Micillo.—Dimelo, no tengas
vergüenza de lo contar á un tan
amigo y compañero tuyo como
yo.
Gallo.—Comence luego de siguir
la tirania y porque tenia sospecha
de mis hermanos yo los degolle y
despues los quemé á ellos y á
mis parientes y aquellos mayores
de la ciudad, que fueron mas de
mill, y despues dobléles el tributo
fingiendo guerras con las
cercanas provincias y grandes
prestamos; mi intencion era
aumentar tesoros para defender
mi misera vida; deleitabame
mucho en cortar cabezas de los
mayores y en robar haciendas de
los menores; hacia traer ante mí
aquellas riquezas; deleitabame en
verlas; en fin, todo este mi deleite
se me convertio en gran trabajo y
pesar, porque como el pueblo se
agraviase con estas sinrazones,
conspiraron contra mi y por
defenderme retrajeme á la
fortaleza con algunos que me
quisieron seguir. Ya estando allí
cercado, yo aun quisiese usar de
crueldad porque inviandome
embajadores de paz los prendí y
los maté y plugo á Dios que por
mi maldad fue echado por fuerza
de allí y fueme acoger con los
lucrenses, que era una ciudad
sujeta á Siracusa, y ellos me
rescibieron muy bien como no
sabian que yo iba huyendo; yo
como hombre habituado á las
pasadas costumbres comence á
robar entrellos (sic) lucrenses las
haciendas de los ricos, tomando
las mujeres hermosas á sus
maridos y sacando las encerradas
doncellas que estaban
consagradas á los templos, y
robaba los templos de todos los
aparejos de oro y plata que habia
para los sacreficios, y con estas
obras vinieronse los lucrenses á
enojar de mi; ¡oh omnipotente
Dios! y qué trabajo tenía en
conservarme en la vida; ¡cuán
temeroso estaba de morir! ni
osaba beber en vaso, ni aun
comer ni dormir, porque en lo uno
y en lo otro temia que me habian
de matar; ¿qué más quieres, sino
que te doy mi fe que con un
carbon ardiendo me cortaba la
barba por no me fiar de la mano y
navaja del barbero, y trabajé por
enseñar el oficio de barbero, á
unas dos hijas que yo tenia,
porque me quemaba con el
carbon que no lo podia ya sufrir?
Despues que por seis años pasé
estos trabajos, no me pudiendo
sufrir los lucrenses echaronme
por fuerza de la tierra, y sintiendo
en paz á Siracusa volvime para
ella, y como de ahi algunos dias
yo volviese á ser peor me
venieron á echar de la tierra jion
(sic) e yo desventurado, corrido y
afrentado, sin poderle resistir me
fue[292] en Corintio destruido por
me guarescer; aqui vine á vevir
en mucha miseria demandando á
mis amigos y enemigos por
limosna el mantinimiento e no lo
querian dar, á que vine á vevir en
mucha miseria y tanta necesidad
que no tenia una capa con que
me defender del frio; en fin, yo me
vi aqui en extrema miseria, tanto
que me vine á enseñar
mochachos á leer y escrebir
porque de aquel salario me
pudiese mantener.
Micillo.—Mas antes yo he oido
decir que lo hacias por ejercitar tu
crueldad castigando los
mochachos con continas
disciplinas, y eras tan
extremadamente cruel que dicen
de ti que en Siracusa una bieja de
muy grandisima edad rogaba á
los dioses continuamente por ti
que te dejasen vivir por muchos
años, y preguntando porqué lo
hacia, pues toda la cibdad
blasfemaba de ti, respondio que
habia visto en su vida larga
muchos señores tiranos en
aquella ciudad y que de contino
sucedia otro tirano peor y que
rogaba á los dioses que tú
vivieses mucho, porque si acaso
habia de suceder otro tan malo y
más peor, que á todos mandaria
quemar juntamente con Siracusa.
Gallo.—¡Oh Micillo! todo me lo
has de decir, que no callarás algo;
bien has visto el trabajo que
tienen los hombres en el mundo
en el reinar y regir mal las
provincias tiranizando los
subditos; mira el pago que los
dioses me dieron por mi mal vivir;
y si piensas que más descanso y
contento tiene un buen rey que
con tranquilidad y quietud
gobierna su reino, engañaste de
verdad, porque visto he que viven
sin algun deleite ni placer; piensa
desde los primeros justos
gobernadores de Atenas é de
toda Asia, Europa, Africa y
hallarás que no hay mayor dolor
en la vida de los hombres quel
regir y gobernar. Si no, preguntalo
á Asalon (Solon) el cual decía que
tanto cuanto más trabajaba por
ser buen gobernador de su
republica tanto y más trabajo y
mal añadia; pero si consideras tú
cuán gran carga echa acuestas el
que de republica tiene cuidado y
aquel que bien ha de regir las
cosas, piensa que no tiene de
pensar en otra cosa en todos los
dias de su vida, sin nunca tener
lugar para pensar un momento en
su propio y privado bien, con
cuánta solicitud procura que se
guarden y esten en su vigor y
fuerza las leyes quel fundó y no
firmó; con cuánto cuidado trabaja
que los oficiales de su republica
sean justos, no robadores, no
coecheros ni sosacadores de las
haciendas de los míseros de
ciudadanos y qué continua
congoja tiene, considerando
que'stá puesto sobre el pueblo
por propio ojo de todos con el
cual todos se han de gobernar,
como piloto de un gran navio en
cuyo descuido está la perdicion
de toda la mercaderia y junto en
el flete del navio va, y tienen gran
cuidado en ver que si en el menor
pecado ó vicio incurre, á todo el
pueblo lleva de si; de otra parte le
combate su mucha libertad y su
mando y señorio para usar del
deleite de la lujuria, del robar para
adquirir tesoros, vendiendo synos
(sic) preturas y gobiernos para
personas tiranas que le destruyan
los vasallos é suditos, lo cual
huye el buen principe
posponiendo cualquiera interese;
¿pues qué soberano trabajo es
sufrir los adúlteros y lisonjeros
que por servirles le cantan
moviendo al buen rey con loores
que claramente ves que en si
mismo no los hay; pues, ¿qué
afrenta rescibe cuando le canta
en sus versos: hice escaramuzas
notables, si nunca entró en batalla
ni pelea, y cuando le procura
importunar trayendo á la memoria
la genología de sus antecesores,
de cuya gloria, él como buen rey
no se quiere preciar, sino de su
propia virtud? Alleganse á esto
los odios, las invidias, las
murmuraciones de los menores,
de las guerras, disenciones y
desasosiegos de sus reinos, que
todo ha de caer sobre él y sobre
su buena solicitud; pues allende
desto qué trabajos se ofrecen en
las encomiendas de las capitanias
y de los oficios del campo, de oir
las quejas de los miseros
labradores que los soldados les
destruyen sus mieses y viñas y
les roban su ganado, que no
basta mantenerlos de balde, mas
que les toman por fuerza las
mujeres y hijas y sin les poder
defender de todo esto. ¿Di,
Micillo, el buen rey que sintirá,
con que sosiego podrá dormir,
con qué sabor comer é que
felicidad ó deleite piensas que
puede tener? Pues ¿qué te
contaré de los caballeros y
escuderos y continos que
comunican en casa del rey y
llevan salarios en el palacio real,
á los cuales como en el mundo no
sea cosa más baja ni más
enojosa ni desabrida ni más
trabajosa ni aun más vil quel
estado del siervo, ellos se precian
de serlo, con decir que tratan y
conversan con el rey y que le
veen comer y hablar y por esto se
tienen por los primeros; en todos
los negocios y horas con una sola
cosa son contentos, sin tener
invidia de alguno, y tratando ellos
la seda y el brocado y las piedras
preciosas menos pueden y curan
de todos los buenos estados del
vevir y de la virtud que
engrandece los nobres y este
dejan por otros, diciendo que les
sea cosa muy contraria el saber;
en esto solo se tienen por
bienaventurados en poder llamar
amo al rey, en saber saludar á
todos conforme al palacio y que
tienen noticia de los títulos y
señores que andan en la Corte y
saben á cuál han de llamar ilustre,
á cuál manifico, á cuál serenisimo
señor; precianse de saber bien
lisonjear, porque esta es la
ciencia en que más se ha de
mostrar el hombre del palacio.
Pues si miras toda la manera de
su vivir en qué gastan el tiempo
de su vida, ¡oh qué confusion y
qué trabajo y qué laberintio de
eterno dolor! oyémelo y cree que
lo dirá hombre expirimentado y
que todo ha pasado por mi sudor
hasta el medio día porque se
fueron acostar cuando queria
amanescer; luego mandan que
esté aparejado un asalariado
sacerdote que muy apriesa
sacrefique a Dios junto á su cama
á la hora de medio día y despues
comenzanse á vestir con mucho
espacio con todas las
pesadumbres y polidezas del
mundo y a la hora de las vísperas
van á ver si quiere comer el Rey;
¡oh qué hacen en palacio!
dispónense á servir á la mesa; á
la hora que ni entra en sabor ni en
sazon se van ellos á comer frio y
mal guisado y luego á jugar con
las rameras ó acompañar al Rey
doquiera que fuere; venida la hora
de la cena tornan al mismo
trabajo y despues que á ellos les
dan de cenar, á la media noche
vuelven al juego y si juega el Rey
ó Principe ó otro cualquiera que
sea su señor, estan alli en pie
hasta que harto su apetito de
jugar se quieren ir á dormir
cuando quiere amanescer. Pues
las camas y posadas de la gente
de palacio, ¿quién te las pintará?
cada dia la suya y tres ó cuatro
echados en una, unos sobre
arcas é otros sobre cofres
tumbados. En cuanto se debe
estimar; ¡oh vida de más que
desesperados! ¡oh Purgatorio de
perpetuo dolor! Pues entre estos
anda un género de hombres
malaventurados que no los puedo
callar; su nombre es truanes
chucarreros, los cuales se precian
deste nombre y se llaman ansi y
pienso que en los decir su trabajo
no merezco culpa si a[ca]so no
me erré. Estos para ser
estimados y ganar el comer se
han de hacer bobos ó infames
para sofrir cualquier afrenta que
les quisieren hacer; precianse de
sucios borrachos y glotones; entre
sus gracias y donaires es
descobrir sus partes vergonzosas
y deshonestas á quien las quiere
ver; sin ninguna vergüenza ni
temor nombran muchas cosas
sucias las cuales mueven al
hombre á se recoger en si; sirven
de alcahuetes para pervertir á las
muy vergonzosas señoras y
doncellas y casadas y aun
muchas veces se desmandan á
tentar las monjas consagradas á
Dios. Su principal oficio es
lisonjear al que tiene presente
porque le dé y decir mal de la
gente publicando que nunca le
dio; y en fin de todos dicen mal
porque otra vez tienen aquel
ausente. Esta es su vida, este es
su oficio, su trato y conversacion
y para esto, son hábiles y no para
mas; de tal suerte que si les
vedase algun principe esta su
manera de vivir por les rescatar
sus ánimas, no sabrian de qué
vivir ni en qué entender, porque
quedarian bobos, necios, ociosos,
holgazanes, inutiles para
cualquier uso y razon, inorantes
de algun oficio en que se
podiesen aprovechar, en este
género de vanidad, trabajando
hechos pedazos por los palacios
tras los unos y los otros confusos
sin se conoscer y al fin todos
mueren muertes viles é infames;
que estos mismos que les
hicieron mercedes los hacen
matar porque en su
malaventurado decir no les trató
bien. Dejémoslos, pues pienso
nuestra reprension poco les
aprovechará; solo una cosa ¡oh
Micillo! podemos de aqui concluir;
que en la vida y ejercicio destos
necios bobos malaventurados no
hay cosa que tenga sabor de
felicidad, mas gran trabajo y
peligro y desventura para si.
Micillo.—¡Oh! Euforbio, ¡oh!
Pitágoras, ¡oh! Dionisio, que no
sé como te nombre, qué
admirables cosas que me has
contado en el trabajo de mandar
reinos y provincias, á tanto que
me has hecho conceder que no
hay estado mas quieto quel mio,
pues en los reyes y los que
comunican en el palacio real
donde paresce estar la
bienaventuranza está tanto
trabajo y desasosiego de cuerpo y
de ánima que casi no parezcan
vivir. Dime agora porque me place
mucho saber mas; despues que
fueste Dionisio ¿qué veniste á
ser?
NOTAS:
[292] En este diálogo está usado fue innumerables veces en el
sentido de fui.
CAPITULO IX
Que pone como fue trasformado

de Dionisio en Epulon el rico y
cuanto trabajo tiene uno en
ser rico y lo que le sucedio.
Gallo.—Mira, mi amo Micillo, yo

no hago caudal en el nombre,
llámame como mas te placerá.
Sabras que despues de poco
tiempo que fui Dionisio vine á ser
un rico de Siria llamado Epulon el
rico, de cuyo desasosiego y
trabajo te quiero ahora decir. Yo
fue hijo de padres muy ricos; yo
ansi por herencia, como por la
gran contratacion sobrepijé en el
poseer muy mayores tesoros que
ellos, por lo cual fue muy
estimado del pueblo y todos me
deseaban servir; hacianme gran
veneracion con gran reverencia;
no habia noble que en estima se
me pensase igualar; tenia
grandes vajillas de plata, vasos
de oro para me servir en el
comer; hacia grandes convites y
banquetes á mis amigos por
hacer gran fama de mi; servianse
con gran aparato de pajes muy
graciosamente ataviados los
manjares; en mucha copiosidad
aquellos potages y salsas en
perfeccion; asalariaba grandes
cocineros examinados en su arte
que supiesen gran diversidad de
los guisados como para un rey;
mientras comia tenia gran
diversidad de música, de cantores
é instrumentos que daban mucho
deleite; bebia las aguas
destiladas y cocidas y los vinos
puestos á infriar, muy
acompañado de juglares y
chocarreros que me daban á los
convidados mucho placer.
Despues de haber comido jugaba
todo el dia grandes cantidades de
moneda por me solazar;
ataviabame muy suntuosamente;
tenia muy poderosos cavallos; iba
á caza de altaneria y de galgos;
mas ¡ay de mi! que Dios sabe con
qué ánimo hacia yo estas
profanidades, que del alma me
salia cada pequeña moneda que
se gastaba, porque si me
esforzaba á lo hacer era por los
que á mi se allegaban por dar de
mi buena fama, que escondido
donde no me podian ver en mi
casa con mis familiares y
apaniguados esforzábame á
pasar con un misero potaje de
miseras lentejas y aunque en él
no habia para todos poder comer,
siempre andaba amarillo y
pensativo como se me gastaba lo
que con tanto trabajo habia
adquerido yendo á las ferias de
todo Egito e Palestina y aun á las
de Grecia por convenir con los
tratantes y mercaderes y con los
deudores á quien con grandes
intereses y usuras yo prestaba mi
moneda; venia por los caminos y
por el mar aventurando mi
persona y hacienda á los cosarios
que me robasen y me quitasen la
vida, sufriendo las crueles
tempestades que cada hora me
ponian en peligro de me perder;
no osaba dar á ningun mendigo
un solo cornado pensando de me
venir empobrecer; pesábame con
grandisimo dolor en pensar que
con la muerte lo habia de dejar. Si
préstamos ó tributos se habian de
dar al Emperador yo habia de ser
el primero; si guerra habia en la
provincia ó que Roma las quisiese
tener yo habia de ir allá y aun
habia de llevar lanzas á mi costa
y mension; en todo esto pasaba
en el campo la misera vida que
pasan los soldados y suelen
pasar en el campo de la guerra.
Temia siempre si mi hacienda que
habia dejado soterrada pensando

Moral Learning Shaun Nichols: Rational Rules: Towards A Theory of

Uploaded by

Copyright:

Available Formats

Moral Learning Shaun Nichols: Rational Rules: Towards A Theory of

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Moral Learning Shaun Nichols: Rational Rules: Towards A Theory of

Uploaded by

Copyright:

Available Formats

Full download test bank at ebook ebookmass.

Rational Rules: Towards a Theory of

CLICK LINK TO DOWLOAD

The Evolution Of Moral Progress: A Biocultural Theory

From Deep Learning to Rational Machines [converted]

From Deep Learning to Rational Machines 1st Edition

Einstein's Unfinished Dream: Practical Progress Towards

Classics of Moral and Political Theory 5th Edition,

Moral Error Theory 1st Edition Wouter Floris Kalf

A Translation Theory of Knowledge Transfer: Learning

The Uses of Delusion: Why It's Not Always Rational to

British Library Cataloguing in Publication Data

For Sarah and Julia

I. RATIONALITY AND RULES

II. STATISTICAL LEARNING OF

III. PHILOSOPHICAL IMPLICATIONS

My ﬁrst book in moral psychology, Sentimental Rules, emphasized the role

rational learning from the evidence. To locate this in contemporary moral

ﬂexible moral learners, with no innate constraints on the kinds of rules

form of rational sentimentalism, where rational learning and emotions both

matters through conversations, discussions in Q&A, and comments on

1.1 Stimuli for probabilistic inference task 18

4.1 Complete list of violations for overhypothesis study 90

Footbridge: Is it permissible for Frank to push a man off a footbridge and

Footbridge-Allow: Is it permissible for Jeff not to pull a lever that would

Bystander: Is it permissible for Dennis to pull a lever that redirects a

Authority dependence: If the teacher didn’t have a rule against hitting,

  :    5

Disagreement: If John and Mark make different judgments about whether

2. A Challenge to Common-Sense Morality

2.1 Debunking Normative Ethics

reasoning. System 2 processes are ﬂexible, domain general, sensitive to new

  :    7

two processes: the moral condemnation is driven by an affective intuition

[T]he consequentialist weighing of harms and beneﬁts is a weighing

tional reactions that are not responsive to rational considerations, we should

2.2 Debunking Metaethics

and the best-known proposals serve as debunking explanations. In his

² In addition, recent work indicates that Bystander is just as emotionally arousing as

  :    9

Integral emotions may contain information that should appropriately

universalism. However, there is a pressing limitation of the study.

3.1 The Many Rationalisms

“Rationalism” is used in several strikingly different ways in philosophy. For

their evidence. I will argue that several key distinctions in common-sense

  :    11

rationality directly to action. Evidentialist approaches to rationality can, of

necessary for responding to the evidence.⁶ Algorithms might be responsive

[T]o be rational is to reason in accordance with principles of reasoning that

  :    13

Marr illustrates different levels of analysis with the example of a lowly

3.3 Rational Processes

We can now recruit Marr’s distinctions to impose greater speciﬁcity in

start by trying to show that a process is rational at the computational level,

  :    15

4.1 Statistical Learning

Some kinds of statistical learning involve extremely sophisticated techniques

A panel of psychologists have interviewed and administered personality

On the basis of this information, thumbnail descriptions of the 30 engin-