Current Research

“Unfair Offers Seem Less Unfair When It is From a Non-human Agent” by Aaron Garvey, TaeWoo Kim and Adam Duhachek

From an economic standpoint, the rational decision in an ultimatum game is to always accept any positive offer because a small monetary reward is better than nothing. However, numerous studies have shown that a concern for fairness is deeply embedded in the human mind and leads people to reject unequal offers that seem unfair (e.g., $10 offered for the self and $90 for the other).The current research proposes a new framework which introduces a novel agent—namely, a non-human, artificial agent. Show MoreWe hypothesize that individual willingness to accept an unfair offer will increase if the offer is made by an artificial agent (vs. human) agent, as fundamental differences in the perceived motives of human and non-human agents attenuate fairness concerns. In support of this hypothesis, we show that, in varying contexts and levels of unfairness, individuals are more willing to accept unfair offers from artificial agents (Studies 1 and 2). To explain our findings, we examine several potential underlying mechanisms, including the following: decreased perception of exploitation intentions, decreased perception of intentionality of an action (i.e., in the context of an algorithm-based offer), cognitive (vs. emotional) reactions to the offer. Additionally, we also examine whether the extent of an artificial agent’s anthropomorphism moderates our findings.

“Cheating on Machines: Consumers Cheat More on Machines (vs. Humans) Due to Reduced Guilt” by TaeWoo Kim, Hye Jin Lee and Adam Duhachek

Building on the burgeoning literature of consumer dishonesty, the current research examines whether consumers’ dishonest behaviors amplify when interacting with non-human artificial agents. We hypothesize that consumers would act more dishonestly when interacting with an artificial (vs. human) agent due to a reduction in anticipatory guilt from engaging in unethical behavior. Show MoreIn support of this hypothesis, we found that consumers are more likely to cheat on artificial (vs. human) agents when an economic incentive for cheating was provided to do so (e.g., e.g., when providing false reasons for a product return leads to the return being free) and that this effect was mediated by the reduced anticipatory guilt associated with the dishonest behavior (Study 1). In an extension of this finding, we hypothesized that consumers would be more likely to disclose guilt-laden personal experiences to an artificial (vs. humsn) agent, as disclosure to an artificial (vs. human) agent feels less emotionally taxing (e.g., less embarrassing). In support of this hypothesis, we found that consumers are more likely to reveal their guilt-laden experiences in general episodic recall tasks (Study 2) and marketing related contexts (e.g., when a consumption experience made them feel guilty) (Study 3) when consumers believed that they were interacting with an artificial (vs. human) agent. We reconcile these two seemingly different findings – that individuals are more likely to be honest about one’s guilt-laden experiences when interacting with artificial agents, and, that individuals are more likely to be dishonest to artificial agents when given with an economic incentive – by attributing both observations to the attenuation attenuation of guilt tendencies when interacting with artificial agents.

“We vs. Them: Rejection by a Non-human Agent Makes People United” by TaeWoo Kim and Adam Duhachek

Artificial intelligence (A.I.) is exerting increasing role in evaluating humans in various context such as loan approval decisions, hiring decisions and legal decisions. Drawing on self-construal theory positing three levels of identity (i.e., individual identity, social identity, human identity), the current research hypothesizes that being rejected by an A.I. (vs. a human) would make a rejected consumer’s human identity more salient, thus leading to reduced perceived self-other distance. Show MoreIn support of this hypothesis, we found that being rejected by an A.I. (e.g., rejected by an A.I. in a job application) increases empathy and perspective ns tendencies of the rejected consumer. These results show that that rejection by an A.I. may increase perceptions of closeness between the self and others because rejection makes rejected consumer’s human identity salient and induces a mindset that tends to embrace others.

“Preference for Human and Non-human Agent in Random Events: Effect of Probability and Outcome Valence” by TaeWoo Kim, Joseph Goodman and Adam Duhachek

Prior research on illusion of control has shown that individuals prefer to make their own choices in a random event (e.g., preferring to choose a lottery number on their own vs. to have it chosen by someone else), believing that they can control the outcome. In the current research, we introduce a novel framework in which we compare human agents with a previously unexamined novel agent in this literature, namely, an artificial agent (e.g., A.I., algorithms, robots). We hypothesize that people will perceive a greater illusion of control in a moderate probability event (i.e., 50% chance of winning) when a human (vs. an artificial agent) is involved.  Show More In the case of a high probability event (i.e., 90% chance of winning), we hypothesize that people will feel stronger certainty for the outcome when an artificial (vs. human) agent is involved because artificial agents are perceived as more likely to translate 90% into a real outcome, whereas human agents are considered to be relatively more error-prone, thus making the low probability negative outcome (i.e., 10% chance of losing) loom larger compared to an artificial agent. In support of these hypotheses, we found that individuals would prefer a human (vs. artificial) agent in a random event (e.g., as an agent who throws the dice) when the probability of the positive outcome is moderate (e.g., 50% chance of winning $100 in a card game) (Study 1). In high probability events (e.g., 90% chance of winning $100 in a card game), however, we found that consumers feel a greater sense of control when an artificial (vs. human) agent is involved in a random event (Studies 2 and 3). We currently seek more evidence of the hypothesized effect in various probabilistic consumer contexts – for example, in receiving medical treatment (e.g., 50% chance of recovery from a disease) or in choosing products with given probabilities of positive or negative outcomes (e.g., 70% chance of a satisfactory restaurant experience).

“Embodiment Effects in Moral Cleansing” by TaeWoo Kim, Adam Duhachek, Pablo Briñol, Spike Lee and Richard Petty

Immoral behavior elicits negative emotions and activates the goal of downregulating negative emotions. Prior research has shown that cleansing the body (e.g., washing hands) helps people attain this goal, owing to the metaphorical association between cleanliness and morality (Lakoff & Johnson, 1980; Lee & Schwarz, 2011; Zhong & Liljenquist, 2006). We challenge this assumption and propose that the effect of physical cleansing on reducing guilt depends on the meanings associated with the cleansing action.  Show More Furthermore, we propose that the emotion-reducing effect of cleansing is stronger when the guilt is elicited by conducting an unethical action (vs. inaction guilt) because the former is more associated with metaphorical associations with contamination and cleansing (e.g., “putting blood on one’s hands”). In Study 1, we showed that, when consumers applied gel to their hands, they were more likely to experience a greater reduction of induced guilt when the gel was framed as “hand sanitizing”—as opposed to having a different meaning, unrelated to cleansing (e.g., handgrip enhancement). Study 2 further demonstrated that this effect emerges only when an actual physical action was present, thus excluding an alternative explanation of semantic priming. In Study 3, it was shown that the effect of physical cleansing on the reduction of guilt emerged when the guilt was caused by an action (i.e., conducting an unethical action) but not when the guilt was caused by inaction (i.e., omitting an ethical action). The current research shows that embodiment effect is driven not by the cleansing action itself, but by the meaning ascribed to the action by consumers.

“Persuasion by Artificial Intelligence Differs along Cognitive versus Affective Routes” by TaeWoo Kim, Adam Duhachek

While some advanced A.I. (e.g., Google’s AlphaGo) can surpass human cognitive intelligence in many domains, their capability to mimic human emotion is still in its infancy. Thus, emotional capability has long been considered a property A.I. lacks when compared to humans. The current research shows that persuasion by an A.I. is more effective when the product is perceived as cognitive (vs. affective) in its nature. In Study 1, persuasion by an A.I. to buy a book was shown to be more effective when the book was framed cognitively rather than affectively.  Show More In Study 2, consumers were found to bet their money on a baseball team that was recommended by an A.I. (vs. human sports analyst) when baseball was framed as a cognitive (vs. emotional) sport. In Study 3, consumers demonstrated higher intention to watch a movie recommended by an A.I. when the movie was described in cognitive (vs. affective) language. However, this effect was attenuated when the same movie was introduced by a human movie expert. In Study 4, we found that the same persuasive message was more effective when a recipient consumer’s mindset was cognitively (vs. affectively) oriented. In examining the underlying mechanism, we found in Study 5 that the effective persuasion from agent-product type matching is due to a heightened attitude certainty created from the matching (vs. mismatching). In Study 6, we found that increasing anthropomorphic characteristics of an A.I. attenuates the matching effect found in previous studies because individuals perceived greater emotional capability from an A.I. that resembles human appearance.

“The Meaning of Consumer Actions Drives Thought Usage in Self Persuasion” by TaeWoo Kim, Adam Duhachek, Pablo Briñol and Richard E. Petty

The current research demonstrates that thoughts can be treated as if they were physical objects, and that the actions performed related to these thoughts and the presumed meaning of those actions determine the impact of those thoughts on evaluative judgments. Across four studies, consumers first wrote either positive or negative thoughts about various consumers’ products and services. Then, consumers performed different actions with those written thoughts. Show More The meanings of these actions were varied to indicate either high validity (e.g., saving, extending, sharing) or low validity (e.g., deleting, hiding, archiving) with respect to their thoughts. We hypothesized and found that performing actions associated with a meaning of validity (vs. invalidity) increased reliance on those thoughts in forming evaluations and behavioral intentions. Furthermore, the validity of those actions’ meanings impacted attitudes by affecting the proposed mediating mechanism (thought confidence). Among other implications, these findings provide the first mediational evidence regarding thought-objectification, extending the work on embodiment, meta-cognition, and consumer evaluation.

“Activation of an Ideal Self Makes Successful Performance Contagious” by TaeWoo Kim, Adam Duhachek and Kelly B. Herd

Contagion beliefs refer to the perception that another individual’s traits can be transferred to the self through direct physical contact with that individual or via a contagious object. Whereas previous contagion research examines contagion effects as a function of the contagion source, we propose that recipient factors may also drive contagion effects. In this view, the same contagion source can produce either positive or negative contagion effects depending upon consumer recipients’ goals. Show More We demonstrate that activation of a goal is a key factor driving contagion effects, leading to a more positive evaluation of a contagion object (Study 1) and enhanced performance in a task related to one’s goal (Study 2), but only when the object was physically touched by a goal-congruent contagion source (Study 3). We find that contagion effects are amplified when consumers are further from their goals (Study 4) and that these effects are attenuated when consumers are in an entity (vs. incremental) implicit theory mindset (Study 5). The implications of these findings for contagion and goal theories are discussed.