Journal of Cleaner Production 285 (2021) 125242

Contents lists available at ScienceDirect

Journal of Cleaner Production

journal homepage: www.elsevier.com/locate/jclepro

Artificial intelligence-enabled environmental sustainability of

products: Marketing benefits and their variation by consumer,
location, and product types
€ rn Frank
Bjo
Faculty of Commerce, Waseda University, 1-6-1 Nishi-Waseda, Shinjuku-ku, Tokyo, 169-8050, Japan

a r t i c l e i n f o a b s t r a c t

Article history: Firms are developing AI-enhanced products (e.g., robots) that can tackle environmental problems
Received 26 August 2020 through autonomous interactions with their surroundings (e.g., removing waste/pollutants, tracking
Received in revised form invasive species) and autonomous learning, which results in improved environmental performance
13 November 2020
characteristics. Such autonomous environmental benefits of products differ from conventional, static
Accepted 20 November 2020
environmental benefits, which derive from pre-purchase processes and design decisions. However, the
Available online 30 November 2020
literature still lacks knowledge of how to use such autonomous environmental benefits to attract new
^as de
Handling editor: Cecilia Maria Villas Bo customers. Therefore, drawing on signaling theory, this study examines the effect of these environmental
Almeida benefits on a consumer’s purchase intent and its variation across types of consumers, locations, and
products. Based on hierarchical linear modeling of 1635 consumer evaluations of AI-enhanced products,
Keywords: this study finds that both static and autonomous perceived environmental benefits influence purchase
Artificial intelligence intent positively. The effect of autonomous environmental benefits is stronger for women than for men
Autonomy and for products targeted at adults rather than children. The effect of static environmental benefits is
Environmental sustainability
stronger for men than women, for products targeted at children rather than adults, for consumers with a
Corporate social responsibility
higher need for cognition, and in locations with a higher perceived environmental well-being.
Green purchasing
Robotics © 2020 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/).

1. Introduction practitioners, scholars, and the public. Engineers have developed

new digital technologies, such as artificial intelligence (AI), to
An environmentally sustainable product contributes less to enhance the environmental sustainability of products. AI refers to
environmental problems than a regular product. This difference the intelligence displayed by advanced machines, as opposed to the
results from environmentally friendly characteristics of its mate- natural intelligence displayed by humans and animals (Poole et al.,
rials, manufacturing processes, distribution processes, disposal/ 1998). It includes capabilities such as the autonomous under-
recycling processes, or product functionality (e.g., low energy standing of the surroundings, learning from experience, decision-
consumption) (Ottman, 2011). Numerous studies report a positive making, implementation of decisions, and advanced communica-
effect of the perceived environmental sustainability of a product on tion with humans and other machines (Russell and Norvig, 2009).
a consumer’s intent to purchase the product (Choi and Ng, 2011; AI may endow products with the ability to tackle environmental
Koller et al., 2011; Nyilasy et al., 2014). Owing to this effect, envi- problems through autonomous actions. For example, firms are
ronmental sustainability tends to increase the profitability of a firm, developing AI-enhanced robots that autonomously clean up houses
despite frequently entailing higher costs (Fraj-Andre s et al., 2009). (e.g., floor, grills, lawns, carpets, air, kitchens, microwave, garbage
Therefore, many firms nowadays strive to enhance the environ- bins, showers, toilets, windows, roofs, pools, excrements of pets,
mental sustainability of their products in order to reap marketing laundry, food recycling), neighborhoods, cities, ponds, lakes, and
benefits and increase their profitability (Herbas Torrico et al., 2018). rivers from garbage, pollutants, micro-plastics, and oil (Abrams,
In recent years, the digital transformation of societal practices, 2018; Chen, 2019; Community Research and Development
business models, and products has aroused the interest of Information Service, 2013; Gerhardt, 2020; Gowan, 2017; Gray,
2019; Knobloch, 2020; Massachusetts Institute of Technology,
2010; Peters, 2019; Sorrel, 2009; Uçar et al., 2020). Other firms
E-mail address: [email protected]. are developing robots that monitor plant health and invasive

https://doi.org/10.1016/j.jclepro.2020.125242
0959-6526/© 2020 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
B. Frank Journal of Cleaner Production 285 (2021) 125242

species (e.g., snakes or fish) (CBS News, 2017; Polverino et al., 2019; customers (Herbas Torrico et al., 2018). By contrast, signaling the-
Rizk and Habib, 2018), robots and AI routines that enhance the ory focuses on the use of sustainability to signal desirable unob-
sustainability of agricultural processes and food production (Di servable characteristics of a product or firm to consumers and thus
Vaio et al., 2020; Kaab et al., 2019; Nabavi-Pelesaraei et al., 2019; also applies to first-time purchases of products (Connelly et al.,
Najafi et al., 2018), AI routines that automate environmental sus- 2011; Spence, 2002). This study draws on signaling theory due to
tainability assessments of products and countries (Carlson and its focus on novel, AI-based technology products, which most
Sakao, 2020; Nilashi et al., 2019) and optimize energy consump- customers have not purchased yet. First, the environmental sus-
tion and distribution (Nizetic et al., 2019; Nosratabadi et al., 2019), tainability of a product signals that the firm offering the product
and robotic vehicles that optimize routes and driving styles to has ethically superior values. Consumers form positive attitudes
minimize their carbon dioxide (CO2) emissions (Alexander-Kearns and intentions toward such products because they identify, and
et al., 2016; Frank, 2018; Nosratabadi et al., 2019). The magnitude thus wish to associate themselves, with these values and because
of this new type of AI-enabled environmental sustainability would they seek to signal to other consumers that they also have these
depend on the post-purchase, autonomous learning of an AI- ethically superior values, which may improve their social re-
enhanced product and its autonomous interactions with its local lationships (Koller et al., 2011; Martínez and Del Bosque, 2013).
surroundings, whereas conventional environmental performance Second, environmental sustainability signals trustworthiness
characteristics are determined by pre-purchase design and process (Martínez and Del Bosque, 2013). As not all quality characteristics of
decisions, which cannot be changed after the purchase (Ottman, a product can be observed before the purchase (e.g., long-term
2011). Thus, this study refers to this novel, AI-enabled type of reliability, detailed functionality), consumers draw upon this
environmental sustainability as autonomous environmental bene- signal of trustworthiness to make inferences regarding unobserv-
fits of a product, whereas it refers to conventional environmental able quality characteristics, which translates into favorable atti-
sustainability as static environmental benefits of a product. tudes and intentions toward the product (Herbas Torrico et al.,
So far, no business-related research has examined the effects of 2018; Martínez and Del Bosque, 2013). Consequently, this present
AI-enabled environmental sustainability on market actors. This study adopts signaling theory (Connelly et al., 2011; Spence, 2002)
study aims to fill this gap in the literature and to identify ways for to develop its hypotheses.
firms to reap marketing benefits from the development of products
with AI-enabled environmental sustainability. Drawing on 2.2. Artificial intelligence and environmental sustainability
signaling theory (Connelly et al., 2011; Spence, 2002), it extends the
literature by comparing the effects of static and autonomous In the field of engineering, several studies address the potential
environmental benefits on a consumer’s intent to purchase an AI- for artificial intelligence (AI) to enhance the environmental sus-
enabled product. Moreover, it examines how the effects of static tainability of products. Likewise, numerous firms are developing
and autonomous environmental benefits vary across consumers, products, where AI enhances the degree of environmental sus-
consumer locations, and product types. These moderators may alter tainability. However, in the field of business, no research appears to
the effectiveness of environmental benefits as signals of unob- examine the effects of AI-based enhancements of environmental
servable product characteristics that trigger purchase intent sustainability on market players’ attitudes and behaviors. To extend
(Herbas Torrico et al., 2018). This study tests the hypotheses with the literature, this study explores the effects of the AI-enhanced
hierarchical linear modeling of 1635 consumer evaluations of AI- environmental sustainability of a product on a consumer’s pur-
enhanced products. chase intent.
To contrast AI-enhanced and conventional types of environ-
2. Conceptual background mental sustainability and thus highlight the differences between
these two concepts, this study divides the different environmental
2.1. The mechanisms linking environmental sustainability and benefits of a product, which together comprise its overall envi-
purchase intent ronmental sustainability, into two groups. First, it defines static
environmental benefits as the environmental benefits that result
After comparing different available products, consumers seek to from pre-purchase design, production, and distribution processes
purchase the product with the highest perceived value, which is the of a product, which cannot be changed or undone after the pur-
perceived gap between benefits obtained and sacrifices incurred chase. For instance, the CO2 emissions during the manufacturing of
(Zeithaml, 1988). Firms aim to maximize the perceived value of a product, the choice of product materials, and the development of
their products by increasing the level and number of benefits that a energy-saving functionalities of a product cannot be undone after
product brings to a consumer’s life, by lowering the price, or by selling the product to consumers. While a part of the post-purchase
both of these strategies (Babin and Harris, 2017). However, unlike environmental impact of a product depends on the extent of its
other product benefits, such as quality attributes, the environ- post-purchase use, the eco-friendly nature of technological features
mental sustainability of a product constitutes a benefit to nature (e.g., whether a car is energy-efficient or not) is determined in pre-
and society, rather than to an individual consumer (Ottman, 2011). purchase development processes. Static environmental benefits
Consequently, environmental sustainability had long been consid- correspond to the traditional notion of environmental sustainabil-
ered irrelevant to consumer behavior. Yet, since the 1990s, con- ity, whose effect on consumer behavior is already known (Choi and
sumer research has identified positive effects of perceived Ng, 2011; Koller et al., 2011; Martínez and Del Bosque, 2013; Nyilasy
environmental sustainability, which may differ from actual envi- et al., 2014). Second, this study defines autonomous environmental
ronmental sustainability (Sen et al., 2006), on consumer attitudes benefits as the ability of an AI-enhanced product to autonomously
and intentions toward products (Choi and Ng, 2011; Koller et al., identify environmental problems, learn and find solutions, and
2011; Martínez and Del Bosque, 2013; Nyilasy et al., 2014). carry out self-determined actions to tackle these environmental
To explain such effects of perceived environmental sustain- problems. While the extent of static environmental benefits is
ability, scholars use multiple theories. Stakeholder theory high- determined by the pre-purchase design, production, and distribu-
lights the use of sustainability by a firm to build goodwill with tion of products, autonomous environmental benefits arise from
stakeholders, such as customers, and is thus more appropriate for post-purchase autonomous interactions between an AI-enhanced
examining the long-term relationship between a firm and its product and its environment, which include learning and
2
B. Frank Journal of Cleaner Production 285 (2021) 125242

decision-making. For instance, a household robot might autono- response. Autonomous environmental benefits may be even more
mously clean up the house and its surroundings from dust, mold, influential than static ones because the consumer has a certain
garbage, and pollutants with tools and devices it purchases and authority over the autonomous (not predetermined and static)
picks up autonomously. Alternatively, it might analyze the con- behavior of an AI-enhanced product, whose actions can thus serve
sumer’s eating habits, identify environmentally friendlier (e.g., as a stronger social signal of the consumer’s own values. For
organic) options, procure these items, and optimize the cooking example, when a consumer directs an AI-enhanced humanoid
procedures to minimize their environmental footprint. household robot to clean up garbage and pollutants in the neigh-
The literature on environmental marketing and business has not borhood, the social environment is likely to interpret these actions
yet addressed autonomous environmental benefits. Drawing on as a signal of the consumer’s own values.
signaling theory (Connelly et al., 2011; Spence, 2002), this study
H1a. Perceived static (non-AI) environmental benefits have a
extends the literature by exploring the effect of autonomous
positive effect on product purchase intent.
environmental benefits on consumer behavior and by comparing it
with the effect of static environmental benefits. Moreover, it ex- H1b. Perceived autonomous (AI-enabled) environmental benefits
amines how these effects vary by the consumer’s gender, need for have a positive effect on product purchase intent.
cognition, location, and evaluated product type. According to
signaling theory (Connelly et al., 2011; Spence, 2002), the influence 3.2. The effects of environmental sustainability types: differences by
of a signal (e.g., the environmental sustainability of a product) consumer
depends on the receiver’s interpretation of the signal (i.e., the
consumer) and on the value of the signal in the receiver’s situation According to signaling theory (Connelly et al., 2011; Spence,
(i.e., location, product context). Fig. 1 provides an overview of the 2002), the influence of a signal, such as the environmental sus-
conceptual framework of this study. tainability of a product, depends on the receiver’s interpretation of
the signal. Since different consumers may differ in their interpre-
3. Development of hypotheses tation of the signal of environmental sustainability, the effects of
static and autonomous environmental benefits on purchase intent
3.1. AI and non-AI types of environmental sustainability: effects on may vary across consumers. Specifically, they may differ between
product purchase intent male and female consumers, whose different social roles affect
their susceptibility to signals of different unobserved characteris-
Drawing on signaling theory (Connelly et al., 2011; Spence, tics of a firm or product (Frank et al., 2014). They may also vary by
2002), scholars argue that the perceived environmental sustain- the consumer’s preference for effortful thinking as signals differ in
ability of a product affects purchase intent positively because it their degree of abstraction and may thus require different degrees
serves as a signal of the trustworthiness and values of the firm of effortful thinking to decode these signals.
offering the product (Herbas Torrico et al., 2018). This signal en- Differences by gender. According to the literature, women are
hances the consumer’s quality perception (Koller et al., 2011; more risk-averse than men and thus more sensitive to signals of
Martínez and Del Bosque, 2013), identification with the brand trustworthiness (Schwartz and Rubel, 2005), also in their pur-
(Martínez and Del Bosque, 2013), and desire to use the product as a chasing decisions (Frank et al., 2014). Moreover, gender roles cause
means of signaling own values to the social environment (Koller women to show a greater desire to signal to their social environ-
et al., 2011). In turn, these mechanisms enhance the consumer’s ment that they adhere to social rules, whereas men have more
intent to purchase the product (Choi and Ng, 2011; Herbas Torrico freedom, or are even socially expected, to sometimes deviate from
et al., 2018; Koller et al., 2011; Martínez and Del Bosque, 2013; social rules to show their audacity and braveness (Holmes, 2013).
Nyilasy et al., 2014). While this argumentation concerns perceived Consequently, the literature reports greater effects of environ-
environmental sustainability in the traditional sense, which this mental sustainability on consumer behavior for women than men
study refers to as perceived static environmental benefits, it may (Lee, 2009; Sudbury Riley et al., 2012; Wang et al., 2018), although
equally apply to the perceived autonomous environmental benefits Mostafa (2007) reports the opposite tendency. This literature fo-
of an AI-enhanced product. Once consumers perceive such benefits cuses only on perceived static (non-AI) environmental benefits.
before the purchase, they likely add them to the sum of environ- However, the argumentation can be extended to perceived auton-
mental benefits expected, which would amplify the signal of omous environmental benefits, which constitute a contribution to
environmental sustainability and the consumer’s resultant society through the actions of an AI-enhanced product and thus

Fig. 1. Conceptual framework and hypotheses.

3
B. Frank Journal of Cleaner Production 285 (2021) 125242

also are a signal of trustworthiness and socially desirable values. In sustainability of a product, depends on its value in the receiver’s
particular, the gender difference in the consumer’s importance situational context (i.e., location, product context). When con-
attached to signaling one’s own values may be even stronger for sumers interpret the environmental sustainability of a product as a
autonomous environmental benefits than for static ones because signal that is more valuable to their situation, they are more likely
the consumer’s authority over the (non-static) actions of an AI- to purchase the product.
enhanced product makes it more likely that the social environ- Consumer location. Regarding the consumer’s location, the
ment attributes these actions to the consumer’s own values. The literature reports differences in the effect of static environmental
expectation of such social recognition would increase the con- benefits on consumer behavior between urban and rural locations
sumer’s motivation to purchase the product in order to signal one’s (Tanner et al., 2004) and between countries (Liobikiene_ et al., 2016).
own values to others (Koller et al., 2011). As an extension, this study explores how the effects of environ-
mental benefits vary by the perceived environmental well-being
H2a. The effect of perceived static environmental benefits on
(i.e., the perceived state of the local environment) at the con-
product purchase intent is stronger for women than for men.
sumer’s location.
H2b. The effect of perceived autonomous environmental benefits In a location with a lower perceived environmental well-being, a
on product purchase intent is stronger for women than for men. consumer may interpret the environmental benefits of a product as
a more important signal because they offer a path for improving the
Differences by need for cognition. While perceived environmental
environmental well-being by purchasing the product. A lower
sustainability serves as a signal of values and trustworthiness
perceived environmental well-being in the consumer’s location
(Herbas Torrico et al., 2018; Koller et al., 2011; Martínez and Del
may thus strengthen the effect of environmental benefits on pur-
Bosque, 2013), the concept (e.g., the relationship between product
chase intent. Contrary to this value mechanism, a more polluted
attributes and global warming) is abstract and difficult to under-
environment may cause the consumer to get used to, and become
stand (Ottman, 2011; Vainio, 2019). Moreover, the environmental
less sensitive to, environmental problems and their solutions (Hu
footprint of production processes and product materials is difficult
and Frank, 2019). This sensitivity mechanism would suggest that
to observe, and a solid understanding thus requires knowledge and
a lower environmental well-being reduces the consumer’s sensi-
contemplation (Ottman, 2011; Sen et al., 2006). Therefore, deeper
tivity to the signal of environmental benefits and thus weakens
thinking may lead consumers to a better understanding of the
their effect on purchase intent.
relevance of static environmental benefits, which is necessary for
To resolve the balance between these two opposing effects, this
interpreting them as a signal of trustworthiness and values. In
study highlights the location where the environmental benefits of a
psychology, a consumer’s tendency for deep thinking is captured by
product materialize. Autonomous environmental benefits materi-
the need for cognition, which reflects the preference for deep,
alize in the consumer’s location, where the AI-enhanced product
rather than simple and less effortful, thinking (Cacioppo et al.,
engages in autonomous actions that alleviate environmental
1984). Thus, this study posits that a higher need for cognition en-
problems. These benefits are more valuable when the consumer’s
hances the interpretation of static environmental benefits as a
location suffers from more environmental problems that the AI-
signal of trustworthiness and values, and consequently strengthens
enhance product can address. Consequently, this study posits that
the effect of perceived static environmental benefits on purchase
the perceived autonomous environmental benefits of a product are
intent. Among the limited research about such a mechanism, one
a more valuable signal in a location with a lower perceived envi-
study supports such a mechanism in analyzing the effects of social
ronmental well-being, where they exert a stronger effect on the
(not environmental) advertising (Yang, 2018), whereas another one
consumer’s intent to purchase the product. By contrast, static
fails to support it in analyzing reasons for environmentally friendly
environmental benefits originate in the pre-purchase design, pro-
food choices (Vainio, 2019).
duction, and distribution phases of a product (Ottman, 2011), which
Contrary to the previous moderating effect, this study posits
mostly take place in a location different from the consumer’s local
that a consumer’s need for cognition weakens the effect of auton-
community. These benefits are thus less valuable for improving a
omous environmental benefits on purchase intent for two reasons.
low perceived environmental well-being in the consumer’s own
First, autonomous environmental benefits originate not in the pre-
location. Consequently, in a location with a lower environmental
purchase phase, but in the use phase of an AI-enhanced product,
well-being, the mechanism of lower sensitivity to static environ-
and are thus easy to observe and comprehend. Second, the primary
mental benefits may outweigh the mechanism of a higher value of
appeal of the autonomous environmental benefits of an AI-
static environmental benefits for improving the environment. Thus,
enhanced product is that these actions are autonomous and
this study posits that perceived static environmental benefits have
liberate the consumer from effortful thoughts and decisions. Thus,
a weaker effect on purchase intent in a location with a lower
autonomous environmental benefits may appeal more to con-
perceived environmental well-being.
sumers with a low need for cognition, who wish to minimize
effortful thinking. H4a. The effect of perceived static environmental benefits on
product purchase intent is stronger in a consumer location with a
H3a. The effect of perceived static environmental benefits on
higher environmental well-being.
product purchase intent is stronger for consumers with a higher
need for cognition. H4b. The effect of perceived autonomous environmental benefits
on product purchase intent is weaker in a consumer location with a
H3b. The effect of perceived autonomous environmental benefits
higher environmental well-being.
on product purchase intent is weaker for consumers with a higher
need for cognition. Product type. While the environmental benefits of a product
serve as a signal of trustworthiness (Martínez and Del Bosque,
2013), the importance of this signal depends on how protective
3.3. The effects of environmental sustainability types: differences by
the consumer is of the intended user of the product. Since humans
situational context
instinctively seek to protect children (Winston, 2011), adult con-
sumers likely attribute a greater importance to signals of trust-
According to signaling theory (Connelly et al., 2011; Spence,
worthiness when purchasing products for children. Thus, this study
2002), the influence of a signal, such as the environmental
4
B. Frank Journal of Cleaner Production 285 (2021) 125242

posits that the effect of static environmental benefits, as a signal of Data were collected across mainland China at firms, public in-
trustworthiness, on purchase intent is stronger when adult con- stitutions, public places, universities, and shopping malls via both
sumers purchase products targeted at children (e.g., toys), than an offline survey and an online survey, which led to 44% of the
when they purchase products targeted at themselves or other responses. Respondents received an incentive valued RMB 30 from
adults (e.g., cars). While this mechanism would also extend to a famous e-commerce platform. After removing missing data, the
autonomous environmental benefits, consumers may interpret the final sample includes 438 respondents, who provided 1635 evalu-
autonomous actions of an AI-enhanced product, which is a machine ations of the four AI-enhanced products. For the purpose of testing
lacking human empathy and childcare instincts, as a threat to the effects of consumer location (H4), the sample covers all regions
children (Wong, 2016). This may weaken the interpretation of of mainland China except for Tibet, whose environment and pop-
autonomous environmental benefits as a signal of trustworthiness ulation have particular features. The sample is distributed evenly
in adult consumers purchasing AI-enhanced products for children. across men and women. It is slightly younger than the population,
Hence, this study posits that the effect of autonomous environ- which matches the greater likelihood of young consumer to pur-
mental benefits on purchase intent is weaker when adult con- chase modern technology products (Frank et al., 2015). Table 1
sumers purchase products targeted at children (e.g., toys) than presents the correlations and descriptive statistics of the vari-
when they purchase products targeted at themselves or other ables. These statistics reveal that the sample consists of consumers
adults (e.g., cars). with average expertise and purchase intentions, who may be
considered representative of regular consumers found in the
H5a. The effect of perceived static environmental benefits on
marketplace.
purchase intent is stronger for products designed for use by chil-
dren than for products designed for an adult consumer’s own use.
4.3. Data validity
H5b. The effect of perceived autonomous environmental benefits
on purchase intent is weaker for products designed for use by Non-response bias. A comparison of early and late respondents
children than for products designed for an adult consumer’s own does not indicate any differences, making non-response bias un-
use. likely (Armstrong and Overton, 1977).
Common method variance (CMV). CMV may bias the conclusions
4. Method of statistical analysis. Lindell and Whitney (2001) provide an
established guideline for estimating the extent of CMV that is
4.1. Measurement tool considered stricter and more accurate than traditional approaches
such as Harman’s single factor test, which this study and most
To measure the variables and prepare for testing the hypotheses others pass. They argue that the smallest correlation between
about the causes of variation in purchase intent, a questionnaire variables in a dataset can serve as an upper bound on CMV. This
was developed for a survey of consumer attitudes toward AI- smallest correlation is .07 in this study and .08 for the dependent
enabled products that are sold on consumer markets and can variable of purchase intent (see Table 1), which implies only a
move when carrying out AI-based decisions. As types of AI-enabled limited extent of possible CMV. Moreover, as another established
products, this study uses autonomous vehicles, robotic pets (for approach to estimating the extent of CMV, this study includes the
child use as required for testing H5), robotic vacuum cleaners, and marker variable of loneliness, which is theoretically unrelated to
humanoid household robots. This diversity ensures the ability to the key variables in the study, as required by Lindell and Whitney
generalize the results beyond specific product contexts. Moreover, (2001). It is measured on a 3-item, 7-point Likert scale (Hughes
obtaining responses on multiple products from the same, rather et al., 2004), which entails higher measurement reliability and ac-
than separate, respondents prevents misinterpreting observed curacy than a scale with fewer items and response points, and
attitudinal differences across products that actually result from fulfills the standard criteria of convergent and discriminant val-
unobserved sample differences (Frank et al., 2014). Moreover, this idity: “I often feel that I lack companionship”/“I often feel left out”/
specific choice of products focuses on AI-enabled products that are “I often feel isolated from others” (Cronbach’s a ¼ .83, average
widely expected to play a role in the future and thus have a high variance extracted (AVE) ¼ .65 > all squared correlations). The
likelihood of predicting effects representative of the future. The seven correlations between this marker variable and the other
appendix lists the construct scales and their literature sources. It reflective variables of the model range from .01 and .05. Five of
also includes the scales of two control variables: product-related them are between .01 and .01, three are negative, four are posi-
expertise and product-related environmental expertise. tive, and six are non-significant. These small correlations and their
distribution around zero imply that this study does not appear to
4.2. Data collection and sample suffer from CMV.
Convergent and discriminant validity. Table 1 shows that all
The data collection targeted China, where environmental sus- multi-item constructs fulfill the criteria of convergent and
tainability plays an important role in order to tackle the severe discriminant validity (Hair et al., 2010): Cronbach’s a > .7, com-
environmental problems of the country (Xu and Lin, 2016). More- posite reliability > .7, AVE > .5, and AVE > squared correlations with
over, since Chinese firms are at the forefront of AI development other constructs. The second-order construct of static (non-AI)
(Allen, 2019), AI-enabled products are more widely available than environmental benefits is based on first-order constructs related to
in other countries that suffer from similar environmental problems. the pre-use (a ¼ .95; AVE ¼ .79), use (a ¼ .98; AVE ¼ .89), and post-
In addition, China exhibits large regional differences in its envi- use (a ¼ .95; AVE ¼ .83) phases of the product life cycle (see ap-
ronmental problems (Xu and Lin, 2016), which provides fertile pendix). These first-order constructs also fulfill the criteria of
ground for testing the role of perceived environmental well-being convergent and discriminant validity. In addition, the fit measures
in the consumer’s location (H4). Consequently, the choice of of a confirmatory factor analysis fulfill the standard acceptance
China may allow for a more reliable testing of the hypotheses than criteria of c2/df < 5, CFI  .95, RMSEA  .07, and upper bound of 90%
would the choice of an alternative country with fewer and less RMSEA confidence interval  .1 (Hair et al., 2010): c2/df ¼ 2.94,
geographically diverse environmental problems and with a lower CFI ¼ .99, RMSEA ¼ .03, upper bound of 90% RMSEA confidence
understanding of AI-enhanced products in the population. interval ¼ .04.
5
B. Frank Journal of Cleaner Production 285 (2021) 125242

Table 1
Correlations and descriptive statistics of constructs.

Correlations

Variables 1 2 3 4 5 6 7 8 9

Consumer
1 Female gender (1: female; 0: male)
2 Need for cognition -.21

Consumer location
3 Perceived environmental well-being -.06 .08

Product
4 Product-related expertise -.16 .14 .11
5 Product-related environmental expertise -.12 .15 .17 .50
6 Child use (1: for child use; 0: for own use) .00 -.01 -.01 -.14 -.12

Environmental sustainability
7 Static (non-AI) environmental benefits (2nd-order construct) -.06 .09 .08 .22 .38 -.13
8 Autonomous (AI-enabled) environmental benefits -.07 .12 .07 .17 .35 -.18 .55

Product adoption
9 Purchase intent -.14 .18 .08 .33 .40 -.20 .37 .38

Descriptive statistics
Mean .55 4.40 4.36 2.74 3.12 .24 3.90 3.93 3.48
Standard deviation .50 1.31 1.46 1.53 1.54 .43 1.29 1.57 1.87
Average variance extracted n/a .67 .85 .93 .92 n/a .63 .82 .92
Cronbach’s a n/a .82 .92 .96 .96 n/a .84 .95 .97

Notes: All correlations |r|  .05 are significant at p < .05 (two-sided). Descriptive statistics for mean score across non-standardized items.

5. Results (Frank et al., 2014, 2015).

Main effects. The results indicate that purchase intent is higher
5.1. Hypothesis tests for women than men and for consumers with a high need for
cognition and a high product-related overall expertise and envi-
Model structure. Table 2 presents the results of the hypothesis ronmental expertise. It is higher for product types targeted at adult
tests. To account for the nested data structure of consumer evalu- consumers, rather than at children (i.e., robotic pets). Both static
ations of up to four product types, the hypotheses are tested using (non-AI) and autonomous (AI-enabled) perceived environmental
hierarchical linear modeling (HLM) with product evaluations at benefits have positive effects on purchase intent, which supports
level 1 and consumers at level 2, whereas the alternative use of the hypotheses H1a and H1b. The effect of autonomous environ-
regression analysis would not properly account for the nested mental benefits is slightly larger, in nominal terms, than the effect
structure of the data. Product purchase intent serves at the of static environmental benefits.
dependent variable. As control variables, the HLM model includes Moderating effects. The effect of static environmental benefits on
the consumer’s gender (1: female; 0: male), need for cognition, purchase intent is larger for men than for women (H2a not sup-
perceived environmental well-being in the consumer’s local com- ported), for consumers with a higher need for cognition (H3a
munity, self-assessment of product-related expertise, and self- supported), for consumers who perceive the environmental well-
assessment of product-related environmental expertise. More- being in their local community as better (H4a supported), and for
over, it controls for whether the product type is primarily targeted products targeted at children, rather than at adult consumers (H5a
at children (1: for child use; 0: for own use), which is the case for supported). By comparison, the effect of autonomous environ-
robotic pets, but not for the other product types. It also includes an mental benefits on purchase intent is larger for women than for
intercept and level-specific error terms. To test the hypotheses, the men (H2b supported), and for products targeted at adult con-
HLM model further includes the consumer’s perception of static sumers, rather than at children (H5b supported). The strength of
(non-AI) environment benefits (H1a) and autonomous (AI-enabled) this effect does not vary by the need for cognition (H3b not sup-
benefits of the product (H1b). In addition, it includes two-way ported) and the perceived environmental well-being (H4b not
interaction terms calculated by multiplying these consumer per- supported). Fig. 2 visualizes the moderating effects. In line with the
ceptions by gender (H2), need for cognition (H3), perceived envi- use of standardized variables in the analysis of Table 2, Fig. 2 uses
ronmental well-being (H4), and product type (H5) after standard deviations from the mean as axis units and designates ± 1
standardizing all variables. The model also includes an intercept standard deviation as high/low values of continuous moderating
and level-specific error terms. According to the pseudo R2 values, variables. The alternative use of a stronger departure from the
the model explains 23% of the variance in purchase intent across mean for high/low values of moderators causes a proportionally
product types for the same consumer and 43% of the variance in stronger variation in the slopes depicted in Fig. 2.
purchase intent across different consumers. As in similar studies,
these values reflect that consumers’ purchasing decisions are based 5.2. Robustness tests and additional analyses
not only on environmental sustainability, but also on other factors
such as product and service quality, price, and brand reputation Quadratic terms. When adding quadratic terms of all continuous

6
B. Frank Journal of Cleaner Production 285 (2021) 125242

Table 2
Effects of perceived environmental benefits on product purchase intent.

Independent variables b
Intercept -.015

Consumer:
Female gender (1: female; 0: male) -.061*
Need for cognition .096***

Consumer location:
Perceived environmental well-being -.002

Product:
Product-related expertise .135***
Product-related environmental expertise .162***
Child use (1: for child use; 0: for own use) -.111***

Perceived environmental sustainability of product:

Static (non-AI) environmental benefits (H1a: þ) .152***
Autonomous (AI-enabled) environmental benefits (H1b: þ) .183***

The effects of environmental sustainability: differences by consumer

Female gender  Static environmental benefits (H2a: þ) -.089***
Female gender  Autonomous environmental benefits (H2b: þ) .063**
Need for cognition  Static environmental benefits (H3a: þ) .059*
Need for cognition  Autonomous environmental benefits (H3b: e) -.018

The effects of environmental sustainability: differences by consumer location

Perceived environmental well-being  Static environmental benefits (H4a: þ) .068**
Perceived environmental well-being  Autonomous environmental benefits (H4b: e) -.009

The effects of environmental sustainability: differences by product

Child use  Static environmental benefits (H5a: þ) .044*
Child use  Autonomous environmental benefits (H5b: e) -.050*

Fit statistics:
HLM pseudo R2 (level 1: product evaluation) .225
HLM pseudo R2 (level 2: consumer) .428
Sample size 1635

Notes: *p < .05; **p < .01; ***p < .001 (two-sided p-values). Hierarchical linear modeling (HLM). Effects of
standardized variables and their interactions.

variables to the analysis, none of these quadratic terms is signifi- environmental benefits (H5a) relate to environmental benefits in
cant, and all hypothesis tests lead to identical conclusions. the use phase.
Static environmental benefits: one first-order construct. When
operationalizing static environmental benefits not as a second- 6. Discussion
order construct based on first-order sub-dimensions, but as
merely one first-order construct (a ¼ .96; AVE ¼ .65), then all 6.1. Short summary
conclusions related to the hypothesis tests remain identical. At the
same time, the confirmatory factor analysis indicates much better This study explores the ability of AI to increase both the
fit for a second-order construct. perceived environmental sustainability of a product and, conse-
All constructs formative. When operationalizing all multi-item quently, a consumer’s intention to purchase this product. To this
measures not as reflective constructs (i.e., factors), but as forma- end, this study compares the effects of autonomous (AI-enabled)
tive constructs (i.e., indices) calculated as an average of their and static (conventional) perceived environmental benefits of a
measurement items, then all hypothesis tests lead to identical product on purchase intent and examines the variation of these
conclusions. effects by type of consumer, location, and product. It finds that both
Sub-dimensions of static environmental benefits. An additional static and autonomous perceived environmental benefits affect
analysis replaced the second-order construct of static environ- purchase intent positively (H1a/H1b supported). The effect of
mental benefits by its sub-dimensions of static environmental perceived autonomous environmental benefits is stronger for
benefits in the pre-use, use, and post-use phases of the product life women than for men (H2b supported) and for products targeted at
cycle. The results indicate that the observed gender difference in adults rather than at children (H5b supported). However, it does
the effect of static environmental benefits (H2a) relates to envi- not vary by the consumer’s need for cognition and by the perceived
ronmental benefits in the use (e.g., low energy consumption and well-being of the environment in the consumer’s location (H3b,
CO2 emissions while using the product) and post-use (e.g., recy- H4b not supported). The effect of static environmental benefits is
cling) phases of the product life cycle. Moreover, the moderating stronger for men than for women (contrary to H2a), for products
effects of need for cognition (H3a) and perceived environmental targeted at children rather than at adults (H5a supported), for
well-being (H4a) on the effect of static environmental benefits both consumers with a higher need for cognition (H3a supported), and
relate to the pre-use phase (i.e., manufacturing and distribution). in locations with a higher perceived environmental well-being
Finally, the observed product differences in the effect of static (H4a supported).
7
B. Frank Journal of Cleaner Production 285 (2021) 125242

Fig. 2. Visualization of moderating effects. Notes: Axis unit: standard deviations from mean. Moderator unit for high/low in the case of continuous variables: ±1 standard deviation
from mean.

6.2. Theoretical implications autonomous environmental benefits. By contrast, it finds the

opposite tendency of a greater effect of static environmental ben-
This study makes several contributions to theory. First, it ex- efits for men than for women, which corresponds to the results
tends signaling theory (Connelly et al., 2011), as a theoretical lens obtained by Mostafa (2007). This might be caused by men’s greater
for explaining the effects of conventional environmental sustain- interest in, and knowledge of, technology products (Frank et al.,
ability (Herbas Torrico et al., 2018), into the new age of AI-enabled 2015), which may translate into a deeper comprehension of the
environmental sustainability. It demonstrates that the integration abstract, unobservable static environmental benefits of a product
of AI into products can boost the level of perceived environmental and thus into a stronger signaling mechanism, whose strength
sustainability and, thus, its effectiveness as a signal that triggers depends on the extent of knowledge held (Sen et al., 2006).
purchase intentions. Hence, AI can benefit nature and marketers Third, this study is the first to explore how perceived environ-
alike. AI-enabled, autonomous environmental benefits appear to mental well-being in the consumer’s location moderates the
have an even stronger effect on purchase intent than do conven- signaling effect of perceived environmental sustainability that
tional, static environmental benefits. The marketing benefits of triggers purchase intentions. Similar to recent findings by Hu and
static environmental benefits are limited by their abstract nature Frank (2019) for non-AI settings, it finds a positive moderating ef-
and by the difficulty for consumers to observe them during design, fect on the effect of perceived static environmental benefits.
manufacturing, and distribution processes in the pre-purchase However, it does not find such a moderating effect on the effect of
phase of a product (Ottman, 2011). These characteristics atten- perceived autonomous environmental benefits. Environmental
uate the effectiveness of static environmental benefits as a signal of pollution may decrease a consumer’s sensitivity to abstract, un-
trustworthiness and values. Hence, static environmental benefits observable environmental benefits as a signal of values and trust-
have a strong effect only on consumers with a high need for worthiness, whereas it does not appear to decrease the consumer’s
cognition, whose deeper thinking helps them to comprehend the sensitivity to environmental actions that take place in front of the
abstract and unobservable environmental characteristics of a consumer’s eyes (i.e., autonomous environmental benefits). In
product. By contrast, autonomous environmental benefits in the studying similar moderating effects of the perceived well-being of
form of autonomous actions that an AI-enhanced product (e.g., the global, not local, environment, Dagher and Itani (2014) find a
robot) carries out in front of the consumer’s eyes are easy to negative moderating effect of perceived static environmental
observe and thus effective as a signal, which boosts their influence benefits, whereas Lee (2009) reports a positive moderating effect,
on consumer intentions. This also ensures that a broader set of but only for female adolescents. In light of such limited evidence,
consumers, including those with a low need for cognition, can the present study lends credence to a positive effect, irrespective of
understand these benefits and respond to them by forming pur- gender.
chase intentions. Fourth, this study is the first to compare the effects of perceived
Second, several studies find a greater effect of perceived static environmental benefits of products targeted at adult consumers
environmental benefits on female consumers than on male con- with those of products targeted at children, for whom adults pur-
sumers (Lee, 2009; Sudbury Riley et al., 2012; Wang et al., 2018). chase such products. Since adults tend to be protective of children,
The present study confirms such a tendency only for perceived they value signals of trustworthiness more in caring for their

8
B. Frank Journal of Cleaner Production 285 (2021) 125242

children (Winston, 2011). Consequently, this study finds that problems. Despite the seemingly apparent connection between
perceived static environmental benefits are more influential for environmental problems and solutions, this relationship may be
products targeted at children. By contrast, perceived autonomous complicated as detailed in the development of H4 and found in the
environmental benefits are less influential for products targeted at counterintuitive results of this study. Future research could disen-
children than for those targeted at adults. This is likely because tangle the sensitivity and value mechanisms underlying this rela-
consumers consider the autonomous actions of an AI-enhanced tionship in non-AI settings. Second, scholars may examine more
product lacking human empathy as a safety risk (Tussyadiah and broadly in non-AI settings how the consumer’s attention to envi-
Park, 2018; Wong, 2016) and may thus be more hesitant when ronmental sustainability differs between purchases of gifts for
purchasing such a product for children, of whom they are protec- others and purchases for consumers themselves.
tive (Winston, 2011), also because children have a low ability to
protect themselves as consumers (Frank, 2012). 6.5. Conclusion

6.3. Implications for managers and public policy makers The integration of AI into products represents an opportunity to
boost the environmental sustainability of these products and,
While managers tend to think of AI functions in products as thereby, to increase consumers’ purchasing intentions and appeal
beneficial for saving a consumer’s time by automating manual to new consumer segments less attracted by conventional envi-
processes (Wong, 2016), this study shows that AI can also lead to ronmental sustainability. Hence, AI-enabled environmental sus-
very different, environmental benefits, which appear to trigger tainability can help firms to build new competitive advantage and
strong purchase intentions in consumers. These AI-enabled, more effectively market their offerings to consumers. At the same
autonomous environmental benefits are more influential than time, this effectiveness varies by the type of consumers and prod-
conventional, static environmental benefits. Moreover, they do not ucts. Compared with conventional environmental sustainability,
suffer from the limited response to static environmental benefits by AI-based environmental sustainability offers a path to appeal more
consumers with a preference for simpler thoughts (Yang, 2018) and to female consumers, which may enable firms to use environmental
by consumers residing in polluted areas. Hence, marketers can use sustainability to more broadly engage consumers across social
them for targeting a broader set of consumers. In addition, the boundaries and secure additional sales while benefitting the
combination of both static and autonomous environmental benefits environment. At the same time, consumers do not appear to
can help appeal to both female consumers, who are more sensitive welcome AI-based environmental sustainability when buying
to autonomous environmental benefits, and male consumers, who products for children, which constitutes a boundary condition in its
are more sensitive to static environmental benefits. However, while use for marketing purposes.
static environmental benefits tend to be effective in products for
children (e.g., organic baby food), autonomous environmental CRediT authorship contribution statement
benefits may scare parents away and may thus be less effective
when targeting parents purchasing products for children. Bjo€ rn Frank: Conceptualization, Methodology, Validation,
Public policy makers and social activists frequently discuss the Formal analysis, Data curation, Writing - original draft, Writing -
perils of AI in controlling people, eliminating people’s jobs, and review & editing, Visualization, Supervision, Project administra-
engaging in emotionless actions that hurt people (Crist, 2019; Kak, tion, Funding acquisition.
2018). Contrary to such negative stereotypes, this study shows that
AI may boost the environmental sustainability of products in a way Declaration of competing interest
that increases consumers’ purchase intentions and, consequently,
also firms’ prospective sales. This would contribute to public policy The authors declare that they have no known competing
goals by increasing firms’ motivations to protect the environment financial interests or personal relationships that could have
and by leading to new employment opportunities at firms offering appeared to influence the work reported in this paper.
AI-enhanced products (Reese, 2019). Moreover, the spread of AI-
enhanced products with autonomous environmental benefits Acknowledgments
would help increase the manpower required to address environ-
mental problems. The author is grateful to all survey respondents and to Yang Yang
and Shan Zheng for their contributions to the data collection. He
6.4. Limitations and directions for future research acknowledges financial support from JSPS KAKENHI (grants no.
16K13396, 19K21707, and 20H01554, Japanese government fund-
A limitation of this study is its focus on a topic of the future, ing) and from Waseda University (grant no. 2019C-152).
which has lower certainty than a description of present consumer
behavior and can only measure intentions, as opposed to actual Appendix A. Supplementary data
behavior in the future. Consumers’ perceptions and preferences
may evolve over time as AI becomes more powerful, reliable, and Supplementary data to this article can be found online at
normal to consumers. Moreover, this study examines only four https://doi.org/10.1016/j.jclepro.2020.125242.
product types. However, it intends to spark a discussion and
encourage follow-up research about hitherto overlooked opportu- References
nities that may arise from AI to improve both the environment and
other valuable aspects of a consumer’s life. Such opportunities can Abrams, M., 2018, May 16. Remote Robot Cleans Trash from Water. American So-
ciety for Mechanical Engineers, New York. https://www.asme.org/topics-
increase the product sales of firms. Aside from this main topic, this
resources/content/remote-robot-cleans-trash-water.
study touches upon two hitherto unaddressed research questions Alexander-Kearns, M., Peterson, M., Cassady, A., 2016, November 18. The Impact of
worthy of future scholarly inquiry. First, the literature does not Vehicle Emissions on Carbon Emissions. Center for American Progress, Wash-
address the relationship between environmental problems in the ington, DC. https://www.americanprogress.org/issues/green/reports/2016/11/
18/292588/the-impact-of-vehicle-automation-on-carbon-emissions-where-
consumer’s location and the consumer’s demand for environmen- uncertainty-lies.
tally friendly products as a possible solution to these environmental Allen, G.C., 2019, February 6. Understanding China’s AI Strategy: Clues to Chinese

9
B. Frank Journal of Cleaner Production 285 (2021) 125242

Strategic Thinking on Artificial Intelligence and National Security. Center for a sectional research designs. J. Appl. Psychol. 86 (1), 114e121.
New American Security, Washington, DC. https://www.cnas.org/publications/ Liobikiene,_ G., Mandravickaite, _ J., Bernatoniene,
_ J., 2016. Theory of planned behavior
reports/understanding-chinas-ai-strategy. approach to understand the green purchasing behavior in the EU: a cross-
Armstrong, J.S., Overton, T.S., 1977. Estimating nonresponse bias in mail surveys. cultural study. Ecol. Econ. 125, 38e46.
J. Market. Res. 14 (3), 396e402. Martínez, P., Del Bosque, I.R., 2013. CSR and customer loyalty: the roles of trust,
Babin, B.J., Harris, E., 2017. Consumer Behavior, eighth ed. Cengage, Boston, MA. customer identification with the company and satisfaction. Int. J. Hospit. Manag.
Cacioppo, J.T., Petty, R.E., Feng Kao, C., 1984. The efficient assessment of need for 35, 89e99.
cognition. J. Pers. Assess. 48 (3), 306e307. Massachusetts Institute of Technology, 2010. A swarm of robots to clean up oil
Carlson, A., Sakao, T., 2020. Environmental assessment of consequences from pre- spills. PLAN, vol. 77. Cambridge, MA. https://sap.mit.edu/article/standard/
dictive maintenance with artificial intelligence techniques: importance of the swarm-robots-clean-oil-spills.
system boundary. Procedia CIRP 90, 171e175. Mostafa, M.M., 2007. Gender differences in Egyptian consumers’ green purchase
CBS News, 2017. "Weird and Wired" Tech helps Battle Invasive Species. New York. behaviour: the effects of environmental knowledge, concern and attitude. Int. J.
https://www.cbsnews.com/news/robots-tasers-technology-join-battle-against- Consum. Stud. 31 (3), 220e229.
invasive-species/. Nabavi-Pelesaraei, A., Rafiee, S., Mohtasebi, S.S., Hosseinzadeh-Bandbafha, H.,
Chen, C., 2019, May 15. 12 Smart Home Gadgets that Practically Clean the House for Chau, K.W., 2019. Comprehensive model of energy, environmental impacts and
You. Insider, New York. https://www.businessinsider.com/best-cleaning- economic in rice milling factories by coupling adaptive neuro-fuzzy inference
gadgets-2018-1. system and life cycle assessment. J. Clean. Prod. 217, 742e756.
Choi, S., Ng, A., 2011. Environmental and economic dimensions of sustainability and Najafi, B., Faizollahzadeh Ardabili, S., Shamshirband, S., Chau, K.W., Rabczuk, T.,
price effects on consumer responses. J. Bus. Ethics 104 (2), 269e282. 2018. Application of ANNs, ANFIS and RSM to estimating and optimizing the
Community Research and Development Information Service, 2013, May 1. Robots parameters that affect the yield and cost of biodiesel production. Engineering
Designed to Clean up Our Streets. Science X. Douglas, Isle of Man: Omicron. Applications of Computational Fluid Mechanics 12 (1), 611e624.
https://phys.org/news/2013-05-robots-streets.html. Nilashi, M., Rupani, P.F., Rupani, M.M., Kamyab, H., Shao, W., Ahmadi, H.,
Connelly, B.L., Certo, S.T., Ireland, R.D., Reutzel, C.R., 2011. Signaling theory: a review Rashid, T.A., Aljojo, N., 2019. Measuring sustainability through ecological sus-
and assessment. J. Manag. 37 (1), 39e67. tainability and human sustainability: a machine learning approach. J. Clean.
Crist, R., 2019, July 13. Amazon and Google Are Listening to Your Voice Recordings: Prod. 240, 118162.
Here’s What We Know about That. CNET. https://www.cnet.com/how-to/ Ni
zeti
c, S., Djilali, N., Papadopoulos, A., Rodrigues, J.J., 2019. Smart technologies for
amazon-and-google-are-listening-to-your-voice-recordings-heres-what-we- promotion of energy efficiency, utilization of sustainable resources and waste
know. management. J. Clean. Prod. 231, 565e591.
Dagher, G.K., Itani, O., 2014. Factors influencing green purchasing behaviour: Nosratabadi, S., Mosavi, A., Shamshirband, S., Kazimieras Zavadskas, E.,
empirical evidence from the Lebanese consumers. J. Consum. Behav. 13 (3), Rakotonirainy, A., Chau, K.W., 2019. Sustainable business models: a review.
188e195. Sustainability 11 (6), 1663.
Di Vaio, A., Boccia, F., Landriani, L., Palladino, R., 2020. Artificial intelligence in the Nyilasy, G., Gangadharbatla, H., Paladino, A., 2014. Perceived greenwashing: the
agri-food system: rethinking sustainable business models in the COVID-19 interactive effects of green advertising and corporate environmental perfor-
scenario. Sustainability 12 (12), 4851. mance on consumer reactions. J. Bus. Ethics 125 (4), 693e707.
Fraj-Andre s, E., Martinez-Salinas, E., Matute-Vallejo, J., 2009. A multidimensional Ottman, J.A., 2011. The New Rules of Green Marketing: Strategies, Tools, and
approach to the influence of environmental marketing and orientation on the Inspiration for Sustainable Branding. Berrett-Koehler Publishers, San Francisco,
firm’s organizational performance. J. Bus. Ethics 88 (2), 263e286. CA.
Frank, B., 2012. The formation of consumer attitudes and intentions towards fast Peters, A., 2019, October 26. The Ocean Cleanup Project Is Now Cleaning Plastic Out
food restaurants. Manag. Serv. Qual. 22 (3), 260e282. of Rivers to Stop it from Getting to the Ocean. Fast Company, New York. https://
Frank, B., 2018. Consumer perceptions of the social and environmental sustain- www.fastcompany.com/90422403/the-ocean-cleanup-project-is-now-
ability of robotic vehicles. Global Environmental Studies 13, 113e117. cleaning-plastic-out-of-rivers-before-it-gets-to-the-ocean.
Frank, B., Enkawa, T., Schvaneveldt, S.J., 2014. How do the success factors driving Polverino, G., Karakaya, M., Spinello, C., Soman, V.R., Porfiri, M., 2019. Behavioural
repurchase intent differ between male and female customers? J. Acad. Market. and life-history responses of mosquitofish to biologically inspired and inter-
Sci. 42 (2), 171e185. active robotic predators. J. R. Soc. Interface 16 (158), 2019, 0359.
Frank, B., Enkawa, T., Schvaneveldt, S.J., Herbas Torrico, B., 2015. Antecedents and Poole, D.L., Mackworth, A.K., Goebel, R., 1998. Computational Intelligence: A Logical
consequences of innate willingness to pay for innovations: understanding Approach, vol. 1. Oxford University Press, New York.
motivations and consumer preferences of prospective early adopters. Technol. Reese, B., 2019, January 1. AI Will Create Millions More Jobs than it Will Destroy:
Forecast. Soc. Change 99, 252e266. Here’s How. Singularity Hub, Santa Clara, CA. https://singularityhub.com/2019/
Gerhardt, N., 2020. 14 Robots You Need to Clean Your House. Family Handyman, 01/01/ai-will-create-millions-more-jobs-than-it-will-destroy-heres-how.
Eagan, MN. https://www.familyhandyman.com/cleaning/robots-that-clean- Rizk, H., Habib, M.K., 2018, October. Robotized early plant health monitoring sys-
your-house/. tem. Annual Conference of the IEEE Industrial Electronics Society 44,
Gowan, J., 2017, April 17. Pollution-eating Robots Could Help Protect the Planet. 3795e3800.
Permaculture Research Institute, The Channon, NSW, Australia. https://www. Russell, S.J., Norvig, P., 2009. Artificial Intelligence: A Modern Approach, third ed.
permaculturenews.org/2017/04/17/pollution-eating-robots-help-protect- Prentice Hall, Upper Saddle River, NJ.
planet/. Schwartz, S.H., Rubel, T., 2005. Sex differences in value priorities: cross-cultural and
Gray, R., 2019, May 29. ORCA: China’s River Cleaning Robot. Dyson On, Malmesbury, multimethod studies. J. Pers. Soc. Psychol. 89 (6), 1010e1028.
UK. https://medium.com/dyson-on/orca-chinas-river-cleaning-robot- Sen, S., Bhattacharya, C.B., Korschun, D., 2006. The role of corporate social re-
929ce62269c4. sponsibility in strengthening multiple stakeholder relationships: a field
Hair, J.F., Black, W.C., Babin, B.J., Anderson, R.E., 2010. Multivariate Data Analysis, experiment. J. Acad. Market. Sci. 34 (2), 158e166.
seventh ed. Pearson, Upper Saddle River, NJ. Sorrel, C., 2009, October 16. Scarab: a Roomba for the Mean City Streets. Wired,
Herbas Torrico, B., Frank, B., Tavera, C.A., 2018. Corporate social responsibility in Boone, IA. https://www.wired.com/2009/10/scarab-a-roomba-for-the-mean-
Bolivia: meanings and consequences. Int. J. Corp. Soc. Responsib. 3 (7), 1e13. city-streets.
Holmes, J., 2013. Women, Men and Politeness. Routledge, New York. Spence, M., 2002. Signaling in retrospect and the informational structure of mar-
Hu, Y., Frank, B., 2019, December 2-4. Green purchasing mechanisms: pollution, kets. Am. Econ. Rev. 92 (3), 434e459.
salience, and values. In: Proceedings of the ANZMAC 2019 Conference. Sudbury Riley, L., Kohlbacher, F., Hofmeister, A., 2012. A cross-cultural analysis of
Wellington. pro-environmental consumer behaviour among seniors. J. Market. Manag. 28
Hughes, M.E., Waite, L.J., Hawkley, L.C., Cacioppo, J.T., 2004. A short scale for (3e4), 290e312.
measuring loneliness in large surveys: results from two population-based Tanner, C., Kaiser, F.G., Wo € lfing Kast, S., 2004. Contextual conditions of ecological
studies. Res. Aging 26 (6), 655e672. consumerism: a food-purchasing survey. Environ. Behav. 36 (1), 94e111.
Kaab, A., Sharifi, M., Mobli, H., Nabavi-Pelesaraei, A., Chau, K.W., 2019. Combined life Tussyadiah, I.P., Park, S., 2018. Consumer evaluation of hotel service robots. In:
cycle assessment and artificial intelligence for prediction of output energy and Stangl, B., Pesonen, J. (Eds.), Information and Communication Technologies in
environmental impacts of sugarcane production. Sci. Total Environ. 664, Tourism 2018. Springer, Cham, Switzerland, pp. 308e320.
1005e1019. Uçar, E., Le Dain, M.A., Joly, I., 2020. Digital technologies in circular economy
Kak, S., 2018, February 7. Will Robots Take Your Job? Humans Ignore the Coming AI transition: evidence from case studies. Procedia CIRP 90, 133e136.
Revolution at Their Peril. NBC News, New York, NY. https://www.nbcnews.com/ Vainio, A., 2019. How consumers of meat-based and plant-based diets attend to
think/opinion/will-robots-take-your-job-humans-ignore-coming-ai- scientific and commercial information sources: eating motives, the need for
revolution-ncna845366. cognition and ability to evaluate information. Appetite 138, 72e79.
Knobloch, C., 2020. 9 Robots that Will Do Your Chores. Scripps Networks, Knoxville, Wang, J., Wang, S., Xue, H., Wang, Y., Li, J., 2018. Green image and consumers’ word-
TN. https://www.hgtv.com/design/remodel/products/9-robots-that-will-do- of-mouth intention in the green hotel industry: the moderating effect of Mil-
your-chores-pictures. lennials. J. Clean. Prod. 181, 426e436.
Koller, M., Floh, A., Zauner, A., 2011. Further insights into perceived value and Winston, R., 2011. Human Instinct. Random House, London.
consumer loyalty: a “green” perspective. Psychol. Market. 28 (12), 1154e1176. Wong, J.C., 2016, September 29. ’This Is Awful’: Robot Can Keep Children Occupied
Lee, K., 2009. Gender differences in Hong Kong adolescent consumers’ green pur- for Hours without Supervision. The Guardian. https://www.theguardian.com/
chasing behavior. J. Consum. Market. 26 (2), 87e96. technology/2016/sep/29/ipal-robot-childcare-robobusiness-san-jose.
Lindell, M.K., Whitney, D.J., 2001. Accounting for common method variance in cross- Xu, B., Lin, B., 2016. Regional differences of pollution emissions in China:

10
B. Frank Journal of Cleaner Production 285 (2021) 125242

contributing factors and mitigation strategies. J. Clean. Prod. 112, 1454e1463. Prod. 8 (1), 79e88.
Yang, C.M., 2018. Effects of message strategy and need for cognition (NFC) on Jawahir, I.S., Dillon, O.W., Rouch, K.E., Joshi, K.J., Venkatachalam, A., Jaafar, I.H., 2006,
consumer attitudes: a case of corporate social responsibility (CSR) advertising. September 11-15. Total life-cycle considerations in product design for sustain-
Open J. Bus. Manag. 6 (3), 714e732. ability: a framework for comprehensive evaluation. In: Proceedings of the 10th
Zeithaml, V.A., 1988. Consumer perceptions of price, quality, and value: a means- International Research/Expert Conference on Trends in the Development of
end model and synthesis of evidence. J. Market. 52 (3), 2e22. Machinery and Associated Technology, Barcelona, Spain, pp. 1e10.
Pelletier, L.G., Legault, L.R., Tuson, K.M., 1996. The environmental satisfaction scale:
a measure of satisfaction with local environmental conditions and government
Further reading environmental policies. Environ. Behav. 28 (1), 5e26.
Venkatesh, V., Morris, M.G., Davis, G.B., Davis, F.D., 2003. User acceptance of in-
Chiou, J.S., 2003. The impact of country of origin on pretrial and posttrial product formation technology: toward a unified view. MIS Q. 27 (3), 425e478.
evaluations: the moderating effect of consumer expertise. Psychol. Market. 20 Venkatesh, V., Thong, J.Y., Xu, X., 2012. Consumer acceptance and use of information
(10), 935e954. technology: extending the unified theory of acceptance and use of technology.
Dragomir, V.D., 2018. How do we measure corporate environmental performance? MIS Q. 36 (1), 157e178.
A critical review. J. Clean. Prod. 196, 1124e1157.
Jasch, C., 2000. Environmental performance evaluation and indicators. J. Clean.

11