Human Communication Research, 2024, 50, 184–193

https://doi.org/10.1093/hcr/hqad034
Advance access publication 24 October 2023
Original Research

Minding the source: toward an integrative theory of

human–machine communication
1,
Eun-Ju Lee *
1
Department of Communication & Center for Trustworthy AI, Seoul National University, Seoul 08826, South Korea
*Corresponding author: Eun-Ju Lee. Email: [email protected]

Abstract

Downloaded from https://academic.oup.com/hcr/article/50/2/184/7329158 by guest on 26 January 2025

According to the computers are social actors (CASA) paradigm, a dominant theoretical framework for research on human–computer interaction,
people treat computers as if they were people. Recent studies on human–machine communication (HMC) and human–artificial intelligence (AI)
interaction, however, appear to focus on when and how people respond to machines differently than to human agents. To reconcile this apparent
contradiction, this study reviews critically the two overarching theoretical explanations proposed and tested in each respective tradition, the
mindlessness account and the machine heuristic. After elaborating on several conceptual and operational issues with each explanatory mecha-
nism, an alternative theoretical model of HMC is proposed that integrates both research traditions and generates predictions that potentially devi-
ate from the dual-process models. Lastly, it is discussed how recent developments in AI technology invite modifications to the current under-
standing of HMC and beyond.
Keywords: artificial intelligence, computers are social actors, human–machine communication, machine heuristic, mindlessness.

Once a sacred realm that belonged solely to human beings, (Fogg & Nass, 1997a). Collectively, these findings led to the
communication is no longer an act that only humans can per- conclusion that computers are “social actors” (CASA), if not
form. Going beyond merely transferring messages between full-blown humans, which elicit reactions normally expected in
humans, machines can now serve as both mediators and com- interhuman encounters.
municators that actively participate in the construction of mes- Unlike earlier human–computer interaction (HCI) research,
sages designed for either mass audience or an individual which is heavily grounded in the “equation” between humans
recipient (Sundar & Lee, 2022). If artificial intelligence (AI) and computers, recent works on human–AI interaction (HAII)
refers to machines that are capable of performing tasks that re- seem to be gravitated toward demonstrating how people are
quire human intelligence (Turing, 1950), then the ability to differently oriented toward AI than they are toward humans,
communicate would certainly be among AI’s key qualifications. for better or worse. Considering that HAII is a natural exten-
With the ever-increasing integration of AI into human com- sion of HCI, the apparent contradiction is rather intriguing,
munication processes, a burgeoning field of research has and even ironic. After all, AI as a more advanced form of tech-
formed that addresses how people evaluate and respond to AI nology can resemble humans more closely than its predecessors
that serves in diverse communicative roles, such as a news re- in both appearance and functionality, rendering its nonhuman-
porter in automated journalism (Cloudy et al., 2021), a fact- ness much less evident. To reconcile this inconsistency, the cur-
checker that tags suspicious information (Banas et al., 2022), rent essay first revisits the notions of mindlessness and machine
a friend one can banter with (Brandtzaeg et al., 2022), or a heuristic, which are the two commonly invoked explanations
personal assistant who corrects misspelled words without for why people respond to computers and AI agents the way
even being asked (Hancock et al., 2020). However, before they do. Specifically, after contemplating some conceptual and
evaluating which agent, AI or human, is more effective in empirical issues with the mindlessness account, I attempt a con-
accomplishing specific communicative goals, one should first ceptual explication of machine heuristic in light of cognitive
ask if people respond any differently to AI than to humans, heuristics and dual-process models. Then, I propose an alterna-
and if so, how and why. tive model of human–machine communication (HMC), which
In fact, long before AI appeared on the horizon, Reeves and integrates both CASA and machine heuristic research tradi-
Nass (1996) reported a series of robust findings in their seminal tions. Lastly, it is briefly discussed how future developments in
book, “The media equation: How people treat computers, tele- AI technology might demand modifications to the current un-
vision, and new media like real people and places,” that people derstanding of HMC and beyond.
treat computers as if they were real people. Despite their aware-
ness that computers are not humans, people consistently apply
a range of social rules and scripts, such as gender stereotypes Mindless social responses to computers?
(Nass et al., 1997), attraction based on similarity (Moon & Arguably one of the most influential theoretical frameworks
Nass, 1996), politeness (Nass et al., 1999), and reciprocity that guided HCI research for the past couple of decades, the

Received: 8 April 2023. Revised: 1 September 2023. Accepted: 8 September 2023

C The Author(s) 2023. Published by Oxford University Press on behalf of International Communication Association.
V
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which
permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited.
Human Communication Research, (2024), Vol. 50, No. 2 185

CASA paradigm has established its robustness in numerous personality, shared membership) or general norms (e.g., po-
empirical investigations. Not as conclusive, however, is why liteness, reciprocity), as documented in the communication or
people emit such seemingly irrational responses, treating life- psychology literature. However, not all social responses are
less machines as if they had gender, personalities, feelings, equally indicative of mindlessness. If one reported more posi-
and intentions. Among several explanations proposed, mind- tive affect, more positive evaluations of the interaction, and
lessness appears to have survived (Lee, 2008, 2010; Nass & more positive regard for the computer when it randomly
Moon, 2000; Sundar & Nass, 2000). Just as people were praised their performance (flattery), rather than provided ge-
more likely to comply with a request that contained neric feedback (Fogg & Nass, 1997b), such reactions might
“because” than the one without, regardless of how legitimate represent a mindless, “click-whirr” response (Cialdini, 2007)
the following reason was (Langer et al., 1978), people mind- driven by the ego-enhancement motive. But what about dis-
lessly process and respond to incoming stimuli, thereby failing counting the flattering computer’s output as less trustworthy?
to adjust their default social reactions while dealing with an In Lee’s (2010) study, for instance, participants found the
asocial being that evinces minimal human-like cues. flattering (vs. nonflattering) computer to be more socially at-
In this view, mindlessness represents a general state of tractive, but were more suspicious of its claims and dismissed
mind, rather than a variable. Boldly put, “individuals’ interac- its suggestions (Experiment 2). Suspecting hidden motives of a

Downloaded from https://academic.oup.com/hcr/article/50/2/184/7329158 by guest on 26 January 2025

tions with computers, television, and new media are funda- flattering person requires a higher degree of mindfulness than
mentally social and natural” (Reeves & Nass, 1996, p. 5), taking the groundless praises at face value, for suspicion is “a
such that the tendency to equate mediated objects and experi- dynamic state in which the individual entertains multiple,
ences with real ones is expected to prevail, regardless of one’s plausibly rival hypotheses about the motives or genuineness
age, education level, or technological proficiency. However, of a behavior” (Hilton et al., 1993, p. 502). Still, one can ar-
as subsequent studies focus largely on whether or not people gue that suspecting a computer’s motives overlooks the fact
exhibit social responses in yet another context with a different that computers have no intention, and hence signals mindless-
social rule, the mindlessness account has not been subjected ness. If so, the only way one can demonstrate mindfulness is
to rigorous scientific scrutiny, but rather assumed as a legiti- to treat flattering and nonflattering computers equally, but it
mate theoretical explanation. Several issues seem to deserve is unclear if such indiscriminate reactions represent a more
our attention. mindful act, compared with either favoring or disfavoring a
computer based on its output. Quite contrarily, one may ar-
Falsifiability of the mindlessness account gue that flattery is a “content cue” that is heeded more care-
First, mindlessness as a psychological state was not separately fully when people adopt a systematic processing strategy
measured, nor was it directly tested. Instead, mindlessness (Chaiken, 1980). In fact, the greater suspicion about and the
was either presumed or inferred from people’s reactions to subsequent dismissal of the flattering (vs. nonflattering) com-
computers. That is, if people are mindful, they should not ap- puter’s suggestions dissipated when the participants were dis-
ply social rules to a computer—given that they did, they must tracted by a secondary task, and hence more mindless (Lee,
have been mindless. It seems analogous to saying that if argu- 2010); that is, flattery made a difference in HMC when peo-
ment quality made a difference in persuasion outcomes, peo- ple were more mindful, which counters the mindlessness
ple must have engaged in systematic processing. However, to account.
validate the causal connection between mindlessness and so-
cial responses to a computer, the most straightforward Underspecified psychological mechanism
method would be to systematically vary the degree of mind- Lastly, the potential variability in the activation of social
lessness and see if it alters the extent to which people treat scripts is not fully engaged. In the CASA framework, social
computers like humans in the predicted direction. responses were conceptualized as built-in, default reactions
In one such attempt, Lee (2008) manipulated the number of that are automatically activated upon encountering human-
tasks imposed on the participants (one vs. two; Experiment 1) like cues (Nass & Moon, 2000). Mindfulness is thus required
and the modality of computer output (speech vs. text; in the application stage to suppress, or “undo,” the activated
Experiment 2) and examined if those cognitively busier with scripts that are deemed irrelevant or inappropriate to the deal-
multitasking or less attentive to the computer output while ings with a computer. After all, to assess whether activated
processing text (vs. voice) became more mindless, and thus, knowledge is usable or not is a controlled process (Higgins,
more likely to exhibit gender-typed responses to computers. 1996) that demands additional cognitive efforts. However,
Although the results supported the mindlessness predictions, some might not exhibit gender stereotypes to a computer, not
even if they had been otherwise, one cannot determine because they have deliberately judged the gender-associated
whether it was because the manipulated stimuli (i.e., the num- social script to be inappropriate and thus unusable in HCI,
ber of tasks, the output modality) failed to induce the state of but because they did not subscribe to gender stereotypes in
mindlessness, or because mindlessness failed to trigger social the first place, regardless of their interactant’s ontological
treatment of computers. Unless we measure mindlessness in- identity (i.e., no activation).
dependently and test these two processes separately, the pre- In sum, while a large volume of research has confirmed the
sumed connection between mindlessness and social responses key proposition of the CASA paradigm that concerns how
to computers remains virtually unfalsifiable. people respond to computers, only few studies (e.g., Johnson
et al., 2004; Lee, 2008, 2010; Xu et al., 2022) have attempted
Monolithic treatment of social responses to validate the mindlessness explanation directly, precisely
Second, it is not entirely clear if “social responses” can be conceptualize what “social responses” are, and/or articulate
treated as a single, monolithic entity. As an umbrella term, so- the process through which mindlessness induces social
cial responses refer to the reactions people naturally show to- responses. Nonetheless, recent works on HMC no longer en-
ward other humans based on specific attributes (e.g., gender, dorse the blanket statement that “individuals’ interactions
186 Minding the source in HMC

with computers . . .are fundamentally social and natural” important, proposed the source of variation (machine heuris-
(Reeves & Nass, 1996, p. 5). Instead, researchers have incor- tic) a priori for empirical validation. Still, several conceptual
porated the machine heuristic (i.e., the extent to which people and empirical questions need to be addressed for machine
attribute machine-like qualities to computing systems) as ei- heuristic to be well integrated with the CASA framework and
ther an explanatory mechanism or a contingent condition for inform an integrative theory of HMC.
proposed source effects (human vs. machine), or a lack
thereof (e.g., Liu & Wei, 2018; Sundar & Kim, 2019; Wang, Heuristic vs. personal beliefs vs. folk theory
2021). As a key construct in the heuristic–systematic model (HSM;
Chaiken, 1980), heuristics refer to “relatively general rules
(scripts, schemata) developed by individuals through their
Machine heuristic: conceptual explication and past experiences and observations” (p. 753). They represent
integration with CASA rules of thumb or mental shortcuts that help reduce cognitive
Apart from some exceptions (e.g., Johnson et al., 2004; Lee load associated with information processing and decision
& Nass, 2002; Morkes et al., 1999), most CASA studies did making (Metzger et al., 2010). An implicit assumption is that
not directly compare human–human interaction with HCI, heuristics are shared widely and available to most people,

Downloaded from https://academic.oup.com/hcr/article/50/2/184/7329158 by guest on 26 January 2025

but instead aimed to replicate the rules of interpersonal inter- such that they are readily activated when people are unable or
action within the HCI context. That is, insofar as people unwilling (or both) to process incoming stimuli thoroughly.
exhibited the same pattern of responses to computers as they As such, if there is considerable variability in how strongly
would to humans, like favoring the computer with similar people embrace and use the machine heuristic in their evalua-
than dissimilar “personality” (Moon & Nass, 1996), it was tions and judgments, then it might be more appropriate to de-
less of a concern whether the degree to which people evince fine it as personal beliefs about machines.
such reactions varies depending on the source. Also, if machine heuristic represents lay people’s under-
As AI comes to play increasingly diverse roles in communi- standing or expectations of how machines perform and oper-
cation processes in lieu of human actors, however, the pendu- ate with what consequences, then it sounds virtually
lum has begun to swing in the opposite direction. Rather than indistinguishable from “folk theories.” Defined as “intuitive,
confirming the “media equation,” much HMC research fo- informal theories that individuals develop to explain the out-
cuses on how the source attribution (human vs. machine) elic- comes, effects, or consequences of technological systems”
its different cognitive, affective, and behavioral reactions (e.g., (DeVito et al., 2017, p. 3165), folk theories are believed to af-
Banas et al., 2022; Cloudy et al., 2021; Jones-Jang & Park, fect how individuals use the systems and respond to their out-
2023; Waddell, 2018). This apparent shift makes sense in puts (Huang et al., 2022). If machine heuristic is measured by
light of the surprise value in scientific research. When com- the degree to which people accept such statements as “As
puters were unmistakably different from humans, it was sur- machines are precise, their prediction will be more reliable
prising and thus worth reporting that people still failed to than humans’,” and “As machines are unbiased, machines’
treat them differently. Contrarily, when “AI agents are devel- guide will be more trustworthy” (S. Lee et al., 2023), which
oped to be more intimate, proximate, embodied, and human- capture not just individuals’ beliefs about key traits of
like” (Jones-Jang & Park, 2023, p. 1), it is surprising that peo- machines, but also those about causal consequences inferred
ple, nonetheless, differentiate between AI and human. In gen- from them, then it becomes indistinguishable from folk
eral, “empirical confirmation of surprising predictions theories.
constitutes a non-trivial advance” (Trafimow, 2013, p. 1).
Despite the diverging orientations, HMC studies still tend What does “machine-like” mean?
to assume that people mostly engage in heuristic (vs. system- As simple schemas or decision rules, heuristics encapsulate
atic) processing when dealing with machines, a relatively evaluative principles, such as “(message) length implies
mindless mode of processing, and examine what specific heu- strength,” “more arguments are better arguments,” and “if
ristics are invoked when people make sense of machine- other people think the message is correct then it is probably
generated messages. In so doing, the machine heuristic, which valid” (Chaiken, 1987, p. 4). Although variations exist, ma-
refers to generalized beliefs or “stereotypes about the opera- chine heuristic is often operationalized by how strongly peo-
tion of machines” (Sundar, 2020, p. 79), was proposed as an ple associate with machines some “machine-like” qualities,
explanation for why people respond differently to an AI agent such as being objective, unbiased, accurate, error-free, un-
than its human counterpart. For instance, those with stronger emotional, unyielding, and reliable (e.g., Banks et al., 2021;
beliefs in machine heuristic were more likely to entrust an AI Cloudy et al., 2021; Sundar, 2020; Waddell, 2018; Wang,
(vs. human) agent with their personal information (Sundar & 2021). What merits note is that some studies (e.g., S. Lee
Kim, 2019), experienced lower levels of emotional involve- et al., 2023; Sundar & Kim, 2019) operationalized the ma-
ment with the AI-authored (vs. human-authored) news article chine heuristic to include an explicit comparison with humans
(Liu & Wei, 2018), and evaluated the news to be less biased (e.g., “Machines are more trustworthy than humans,” “When
and more credible when the uncivil comment section was machines perform a task, the results are more objective than
moderated by a machine (vs. human) agent (Wang, 2021). when humans perform the same task”). If machine heuristic
The introduction of machine heuristic as a user characteris- hinges on what separates machines from humans, then stron-
tic that systematically alters the processes and outcomes of ger beliefs in machine heuristic will inhibit, rather than facili-
HMC represents a refinement of the CASA paradigm. That is, tate, social responses to computers, directly challenging the
while inheriting the key proposition that people are generally mindless explanation for the CASA paradigm.
mindless when interacting with machines, and thus prone to Moreover, some researchers included specific behavioral
rely on heuristics, recent studies demonstrated that not all outcomes that would reasonably follow certain beliefs about
people treat humans and computers equally, and more machines as part of the operational definition of machine
Human Communication Research, (2024), Vol. 50, No. 2 187

heuristic (e.g., “Machines can handle information in a secure HSM acknowledges that both heuristic and systematic proc-
manner, therefore it is generally okay to disclose my private essing occur concurrently (Chaiken et al., 1989), that does
information [e.g., card number, address, phone number, etc.] not explain why machine heuristic exerts greater influence on
to them.”; S. Lee et al., 2023). Apart from its double-barreled communication outcomes in the systematic, rather than heu-
nature, if people indeed share more private information with ristic, processing mode. Moreover, if peripheral and central
the AI agent (behavioral outcome) because they believe “it is processing represent the low and high ends of the elaboration
generally okay to disclose their private information to continuum, as per the elaboration likelihood model (ELM;
machines,” it seems dangerously close to a tautology. If ma- Petty & Cacioppo, 1986), the influence of machine heuristic
chine heuristic were to serve as an explanation for why people should be lessened as people process information more effort-
behave toward computers and humans differently, then it fully (Petty et al., 1981). As such, an empirical test is in order
would be more appropriate to separate (a) individuals’ beliefs that examines in which processing mode people rely more
about inherent characteristics of machines and (b) specific heavily on machine heuristic, and thus differentiate between
actions that stem from such beliefs, and define machine heu- machines and humans.
ristic strictly in terms of the former.
Lastly, if machine heuristic refers to generalized beliefs or Machine heuristic as mediator vs. moderator

Downloaded from https://academic.oup.com/hcr/article/50/2/184/7329158 by guest on 26 January 2025

“stereotypes” about what machines are like and how they Another remaining question concerns the theoretical role that
perform (Molina & Sundar, 2022; Sundar & Kim, 2019), its the machine heuristic plays in predicting and explaining peo-
specific content might need to be updated, as people’s expect- ple’s responses to machines (vs. humans) (see Bellur &
ations of machines evolve with technological advancements Sundar, 2014, for a thorough discussion on “heuristics-as-
over time. For instance, if friendly and personable social chat- variables,” p. 119). Similar to other heuristics, machine heu-
bots that convey unwavering support and empathy become ristic is considered to be stored in individuals’ memory and
widely popular, then the currently commonplace notion that “cues on the media interface suggesting a machine source will
machines are “unyielding,” “unemotional,” and “cold” trigger this heuristic, which in turn will shape perceived qual-
(Molina & Sundar, 2022; Sundar, 2020) may lose its ground. ity and credibility of media content as well as the entire user
Likewise, when the generative AI’s ability to create hallucina- experience” (Sundar, 2020, p. 80). Empirical investigations,
tion by providing utterly fabricated information is repeatedly however, seem to diverge in how such processes are modeled
highlighted in the media coverage (Weise & Metz, 2023), and tested.
people might realize that the apparent objectivity of algo- On the one hand, some studies (e.g., Cloudy et al., 2021;
rithms is only skin-deep and AI is not as accurate or reliable Waddell, 2018, Study 2) proposed that machine heuristic
as they once believed. In both cases, the very definition of mediates the effects of source (human vs. machine; IV) on the
“machine-likeness” would need to be revised, but it is unclear users’ cognitive, affective, and behavioral reactions (DVs;
if the current conceptualization of machine heuristic allows Figure 1A). Two conceptual issues merit note. First, it is one
such updates. If it does, can we still call the machine heuristic thing that the exposure to machine agency cues activates the
a heuristic, when the if–then relationship it postulates (e.g., if machine heuristic, and it is another that the source cue affects
a machine wrote the news article, it is unbiased) is in flux? how strongly one endorses the machine heuristic, as a media-
tion model implies. Put differently, there is no inherent reason
Does the machine heuristic operate heuristically?
why people should become more (or less) receptive to the ma-
To confirm that the machine heuristic indeed guides people’s chine heuristic, simply because they encountered a machine
reactions to machines as a mental shortcut, it should be em- agency cue, when the sheer act of asking them about machine
pirically demonstrated that the effects of machine heuristic on heuristic would likely activate their beliefs about machines in
subsequent judgments and behaviors are amplified when peo- the human condition as well. Instead, the degree to which
ple engage in heuristic rather than systematic processing. Put people agree that machines are fair, objective, and error-free
differently, machine heuristic should become more influential is likely to vary as a function of the machine agent’s preceding
when people lack either cognitive resources (e.g., due to multi- performance. After reading a well-written news article osten-
tasking, distraction, insufficient prior knowledge) or the moti- sibly authored by AI, for instance, people would accept the
vation to process the communication systematically (e.g., due machine heuristic more readily (à la availability heuristic;
to low personal relevance or interest), or both (Chaiken et al., Tversky & Kahenman, 1973), as compared with those who
1989). However, heuristic processing was often taken for read a human-authored article. To rule out this potential con-
granted as a default mode of information processing (see the founds, one should examine if exposure to a machine agent’s
MAIN model, Sundar, 2008; Sundar et al., 2019), and the
very fact that beliefs in the machine heuristic served as a sig-
A
nificant moderating variable of source effects was interpreted
Source Type Responses
as evidence for heuristic processing (e.g., Banas et al., 2022; (Human vs. Machine)
Machine Heuristic
Cognitive Affective Behavioral
Banks et al., 2021; S. Lee et al., 2023; Sundar & Kim, 2019).
If people rely more on the machine heuristic while they en-
gage in heuristic (vs. systematic) processing, then it follows B
that the more mindless, the more likely people would treat Machine Heuristic
humans and computers differently—a prediction that directly
contradicts the mindlessness explanation for CASA.
Source Type Responses
Alternatively, if the effects of machine heuristic on individu- (Human vs. Machine) Cognitive Affective Behavioral
als’ responses to machines are amplified when people are
more mindful, then we should probably reconsider its status Figure 1. (A) Machine heuristic as mediator of source effect and
as a heuristic, in light of the dual-process models. Although (B) machine heuristic as moderator of source effect.
188 Minding the source in HMC

poor performance still induces a stronger endorsement of the authenticity model of computer-mediated communication
machine heuristic than does its human counterpart’s. (Lee, 2020), authenticity of communication consists of three
Second, it is unclear why beliefs in machine heuristic (medi- subcomponents: how well the claimed identity matches the
ator) should affect the evaluative outcomes (DVs) in both hu- source’s real identity (authenticity of source, p. 61), how
man and machine conditions alike. By definition, the machine truthfully a message represents its object (authenticity of mes-
heuristic captures people’s expectations and beliefs about sage, p. 62), and how closely people feel they are a part of ac-
machines. If so, beliefs in machine heuristic should be either tual interaction (authenticity of interaction, p. 63). In the
(a) irrelevant to the evaluations of a human agent or (b) in- authentication process, although the model does not use the
versely associated with them, if people automatically compare term, self-confirmation heuristic can serve as a super-heuristic
machines with humans (see Lee et al., 2022 for how a human that precedes and guides the operation of other heuristics, if
agent’s unsatisfactory performance enhances individuals’ rat- ever. Specifically, “when an incoming message deviates from
ings of AI). Either way, the association between the mediator individuals’ expectancy, which is formed based on general
(machine heuristic) and the DVs should vary depending on knowledge, schemas, social scripts, stereotypes, and their past
the agency type (machine vs. human), but the simple media- experiences,” (p. 64) they might simply dismiss the communi-
tion model fails to uncover this interaction. In fact, studies cation as false or inauthentic, and the authentication process

Downloaded from https://academic.oup.com/hcr/article/50/2/184/7329158 by guest on 26 January 2025

found that beliefs in machine heuristic significantly altered the ends immediately. As the sufficiency principle of HSM postu-
evaluations of an AI agent’s performance, but not a human’s lates (Chaiken et al., 1989), however, under some circumstan-
(e.g., Lee et al., 2023b; Sundar & Kim, 2019). ces where the accuracy motivation is high, people are willing
On the other hand, others highlighted the variability in peo- to take further steps, “by utilizing available cues and engaging
ple’s endorsement of machine heuristic as a contingent condi- in different cognitive processes” (p. 64).
tion for the source effect to occur. In this scenario, the machine Unlike HSM and ELM, which posit that people rely more
heuristic moderates the degree to which users differentially re- on heuristics and peripheral cues when they lack the motiva-
spond to machine and human agents (Figure 1B). Although it tion and/or the ability to process messages thoroughly, how-
was not framed this way, the moderating effect of machine heu- ever, the authenticity model proposes that it is when
ristic can be deemed as a special case of a confirmation bias— individuals are more willing to expend extra cognitive efforts
that is, those who believe that AI in general is more objective, to authenticate the communication that they take various cues
accurate, and reliable would think more highly of and respond and associated heuristics into account. In fact, while some
more favorably to a specific AI agent’s performance, be it AI dual-process studies differentiated message (e.g., argument
journalist (Lee et al., 2023b), AI content moderator (Wang, quality) and nonmessage cues (e.g., expertise and likability of
2021), or AI travel agent (Sundar & Kim, 2019). In contrast, source) (e.g., Chaiken, 1980; Petty et al., 1981; see Petty
those who do not espouse such beliefs would either treat et al., 1999 for how this content-based partition should not
machines and humans similarly or even evaluate machines be equated with the conceptual distinction between arguments
more negatively than their human counterparts. and cues), the authenticity model refers to both message and
nonmessage features as “authenticity markers” (Lee, 2020, p.
69; also see Kruglanski & Thompson, 1999) that people may
When do people mind the source? turn to when their expectancy has been violated. In this view,
As discussed above, recent studies on HMC that utilize the while “counting on heuristic cues (e.g., the number of likes)
machine heuristic inherit the key proposition of the CASA while ignoring more substantial information (e.g., argument
paradigm by assuming heuristic (mindless) processing as a de- quality) may signal cursory information processing,” “if only
fault mode, but they also depart from it by challenging the heuristic cues are available . . . utilizing such cues may reflect a
universality of social responses to machines. More important, more effortful cognitive process than turning a blind eye to
machine heuristic presupposes, either explicitly or implicitly, them” (p. 65).
that people associate with machines some unique attributes Recent empirical evidence seems to support this prediction.
that distinguish machines from humans. Such conceptualiza- For instance, Lee et al. (2023a) found that communication
tion leads to the prediction that the more mindless, the more channel (a nonmessage cue) exerted greater influence on indi-
likely people rely on the machine heuristic, thereby exhibiting viduals’ acceptance of a persuasive health message among
greater differentiation between human and machine agents. those more interested in health. Specifically, participants in-
This directly counters the mindlessness account of CASA. ferred higher levels of ulterior motives (i.e., less authentic)
Despite such differences, both traditions assume that people from a medical reporter’s newspaper column than from his
are constantly mindful of the source. It is how people treat dif- Facebook post with identical content, which lowered their in-
ferent sources (human vs. machine), either similarly or differ- tention to follow the recommended behavior. But such a ten-
ently, that these accounts diverge. However, the extent to dency was stronger among those more interested in health,
which people mind the source, whether it is a human or a ma- who presumably processed the health messages more
chine, might vary as a product of situational and dispositional carefully.
factors, and more important, in a way that deviates from Directly germane to HMC, studies have also reported the
what classical dual-process theories would predict. significant effects of source type (human vs. AI) when individ-
uals are more motivated to process information systemati-
Self-confirmation as super-heuristic cally. For example, after reading a news article attributed to
Perhaps better known as confirmation bias, Metzger and either algorithm or a human journalist, participants were
Flanagin (2013) refer to the “tendency for people to view in- more likely to rate the article credibility differently depending
formation as credible if it confirms their preexisting on the source, when it lacked objectivity (i.e., no source attri-
beliefs and not credible if it counters their existing beliefs” as bution, value-laden words like “thankfully”), which would
“self-confirmation heuristic” (p. 215). According to the have triggered message scrutiny (Tandoc et al, 2020).
Human Communication Research, (2024), Vol. 50, No. 2 189

Similarly, Lee et al. (2022) examined how people evaluate the Expectancy Confirmed?
AI (vs. human) moderator of user comments sections and
Yes No
found that participants were more suspicious about the AI
moderator’s ulterior motives, but only when the remaining
comments were counter-attitudinal or when no explanation Human = Machine Task-Relevant Machine Heuristic?

was provided for deleted comments. When the remaining Yes No

comments were mostly proattitudinal or when reasons were
provided for why some comments had been removed, partici- Valence? Human = Machine
pants’ suspicion of ulterior motive remained unaltered,
whether the moderator was AI or a human.
Positive Negative
An integrative theory of HMC: machine heuristic as
Assimilation Machine > Human Machine < Human
conditional moderator
Taken together, the assumption that people mind who or Contrast Machine < Human Machine > Human

Downloaded from https://academic.oup.com/hcr/article/50/2/184/7329158 by guest on 26 January 2025

what the source is, even if they fail to adjust their responses
Figure 2. Machine heuristic as conditional moderator of source effect.
accordingly, might need to be revisited. People might not
care, unless they have a reason to consider who authored a
news article, who checked the facts, who composed a music, and/or opinions, however, they would consider additional
and the like. When the message violates their expectations, factors, like the source, to make sense of the unexpected out-
and thus motivates them to mind the source among other fac- come. With an AI fact-checker, their prior beliefs about AI
tors, then their beliefs about the source, either categorical then guide how they respond to the expectancy-disconfirming
(e.g., machine, AI) or individuated (e.g., ChatGPT, Siri), will message. If people assimilate their responses to their existing
set in and guide the subsequent judgments. Such beliefs are stereotypes about AI, then those holding more positive
likely to form based on one’s personal experiences with stereotypes will be more likely to accept the disconfirming
machines, observations of others’ experiences, and media por- fact-check verdict. Consequently, the confirmation bias in the
trayals of machines, to name a few. What is more, in order acceptance of corrective information will be attenuated in the
for such beliefs to shape evaluative outcomes, they should be AI, rather than human, condition (Figure 3A). In contrast,
relevant to the task at hand. For instance, one’s beliefs about those with more positive beliefs about AI might exhibit even
AI’s fairness might not affect how people evaluate a social more negative reactions to the AI agent’s opinion-challenging
chatbot’s friendliness. verdict due to a negative expectancy violation, thereby ampli-
There are two opposite ways in which one’s prior beliefs fying the confirmation bias in the AI condition (Figure 3B).
about machines bias their reactions to a machine’s perfor- The current model calls for three empirical considerations.
mance that has violated their expectations: assimilation and First, we should rethink what a failed manipulation of source
contrast. First, people might evaluate the machine agent’s out- actually means. When study participants do not recall
put in a belief-congruent manner, such that those holding whether the message was written by algorithm or a person,
more positive beliefs about how machines operate (e.g., being their data are usually discarded (e.g., Tandoc et al., 2020).
fair, objective, accurate, and unbiased) might be more recep- However, when a sizeable proportion of participants turned
tive to the machine agent’s expectancy-disconfirming act, out to have paid little attention to the source, we should per-
when compared with those holding no such beliefs. Albeit not haps ask how mindful people normally are of the message
limited to the cases wherein expectancy violations occurred, source in real life, instead of artificially exaggerating the
studies reported that those with stronger beliefs in machine source label so it will rarely go unnoticed. Second, it is impor-
heuristic (i.e., machines are objective, unbiased, accurate, tant to distinguish between (a) relying solely on the source cue
error-free) were more likely to accept an AI fact-checker’s (AI vs. human) with disregard for the message content and (b)
truth verdicts (Banas et al., 2022) and rated a news article to relying on both message and source cues, for they each repre-
be less biased when the uncivil comment section was moder- sent qualitatively different cognitive processes. By establishing
ated by a machine (vs. human) agent (Wang, 2021). that there is no difference in message elaboration, one can
Alternatively, one’s prior beliefs about machines might serve conclude that the greater source effect is not a sign of heuristic
as an anchor against which the current event is evaluated. In or mindless processing, but a product of more effortful proc-
such a case, those holding a positive machine heuristic will re- essing that took all the available cues into consideration.
spond more negatively to an AI agent’s underperformance, Third, it remains to be tested in which direction prior beliefs
which fell short of their inflated expectations. The finding about machines bias people’s reactions to negative expectancy
that a news article lacking objectivity was penalized more se- violations by AI agents, assimilation or contrast, and more-
verely with lower credibility ratings when attributed to algo- over, if it varies, what factors account for such variation. For
rithm than to a human journalist (Tandoc et al., 2020) might instance, when one’s beliefs about AI are firmly grounded in
suggest a negative expectancy violation, if we can assume that their direct experiences, as opposed to remote observations or
participants overall believed that machines are objective, bal- hearsay (i.e., belief certainty), assimilation might be more
anced, and accurate. Figure 2 summarizes these processes. likely to occur.
Take an AI fact-checker, for example. When its verdict con- At the same time, some limitations should be noted regard-
firms an individual’s existing beliefs, attitudes, and/or opin- ing the proposed model’s generalizability. First, not all com-
ions, they will accept it as is, without considering who munications may allow the binary classification of either
rendered the verdict—AI or human (i.e., no main effect of the confirming or disconfirming expectancy. People might not
source type). When the verdict counters their beliefs, attitudes, even have any prior knowledge or experience to form solid
190 Minding the source in HMC

More important, do we need a theory exclusively about

HMC or can it be subsumed within the purview of a general
communication theory?
In attempts to address these questions, the current essay
reviewed how mindlessness and machine heuristic have been
conceptualized and operationalized in the literature and
raised some questions about their viability as a theoretical ex-
planation for CASA in particular, and for HMC in general.
After identifying potential incompatibilities, an alternative
model informed by the authenticity model of computer-
mediated communication (Lee, 2020) was proposed to recon-
cile the two research traditions and develop an integrative
theory of HMC. Although the authenticity model was devel-
oped for human–human communication via computer, it may
also help explain human communication with computer by

Downloaded from https://academic.oup.com/hcr/article/50/2/184/7329158 by guest on 26 January 2025

integrating the source type (human vs. machine) as a variable.
With the seamless integration of AI technology into numer-
ous applications and services, it has become critical to ensure
transparency concerning the specific roles and functions of AI
from the ethical standpoint, but less clear is how users process
such information, and with what consequences. Both CASA
and machine heuristic studies assume that people are mindful
of the source, even if they often fail to adjust their expecta-
tions and reactions accordingly. However, recent findings
(e.g., Lee et al., 2022; Tandoc et al., 2020) suggest that people
may not be sufficiently mindful of the source, unless they need
to. Insofar as they enjoy the conversation, obtain useful infor-
mation from the news article, and appreciate the song they lis-
Figure 3. (A) Assimilation effect of positive machine heuristic and ten to, why should they care who does the talking, wrote the
(B) contrast effect of positive machine heuristic.
news, and composed the song? After all, we are cognitive
misers (Fiske & Taylor, 1984). If so, minding the source, as a
expectations about the subject of a message. In such cases, cue, and relying on associated cognitive shortcut (i.e., ma-
factors known to facilitate message elaboration, such as the chine heuristic) to make judgments and decisions, which the
accuracy motivation or personal relevance, are likely to deter- ELM has considered as an indication of less mindful process-
mine on which route their message processing will take place. ing, may in fact reflect rather effortful and elaborate process-
Second, the current model is better suited for communica- ing of communication.
tion contexts where the source salience is relatively low. Even Moreover, recent technological breakthroughs seem to
though people are told who/what the source is, reading a by- challenge the CASA’s key premise that it is unreasonable and
line of a news article (“authored by algorithm”) is not the irrational to treat computers like humans. For instance, we
same as interacting with a chatbot in terms of how vividly are witnessing an increasing number of virtual humans who
people feel the machine’s presence. As such, the present model look like a human, dance like a human, and talk like a human
might be more useful in explaining one-to-many, rather than (https://www.virtualhumans.org/). Why then should follow-
one-to-one, communication contexts (i.e., AI Curator and AI ing a virtual influencer be different from admiring BTS or
Creator; Sundar & Lee, 2022). When the agent type is suffi- BLACKPINK? When conversing with a social chatbot who is
ciently salient to begin with, individuals’ beliefs about AI will always there whenever I need him/her and listens to whatever
set in instantly, thereby shaping communication processes I want to vent without worrying about a leak (Brandtzaeg
and outcomes. For instance, the agency type might influence et al., 2022), disregarding the agent’s ontological identity can
the degree of message elaboration, such that people might be a fairly rational choice to gratify one’s hedonistic needs,
process a message more effortfully when it is attributed to AI maximizing the pleasure and emotional support one can de-
than a human, due to the novelty. If so, content cues will exert rive from the “friendship.” In the age of multiple identities
stronger influence when AI has authored the message. that are in constant flux, it may not be too different from en-
gaging with a role-playing partner in the metaverse, wherein
users are fully aware that the avatars’ appearance and person-
Looking ahead and moving forward ality may not match the “real” person behind the cartoon
To develop a theory is to answer the fundamental “why” character, and yet, suspend their disbelief to get fully im-
question about observed regularities in the focal phenomenon mersed in the virtual world and enjoy the experience.
of interest (Berger et al., 2010). As such, to theorize HMC, Once people mind the source, then their beliefs about
one needs to identify consistent patterns in people’s reactions machines in general will direct how the source type affects
to machines, propose the mechanisms to explain why such their subsequent reactions. Thus far, studies employing the
patterns exist, and put them to empirical tests for validation. machine heuristic tend to focus on machines’ agentic, rather
Can mindlessness and/or machine heuristic serve as the core than communal, qualities of largely positive connotations,
mechanism that predicts and explains the processes and such as being objective, accurate, precise, and error-free (see
effects of HMC? Are there any alternative explanations? Molina & Sundar, 2022 for an exception). While those are
Human Communication Research, (2024), Vol. 50, No. 2 191

undoubtedly core traits people seem to associate with Data availability

machines at the moment, more systematic approaches are No data are linked to the current article.
called for, if we are to use this construct to account for a wide
variety of HMC beyond task-oriented domains. Perhaps a
more comprehensive set of adjectives people use to describe References
machines can be gleaned inductively from a large-scale survey Banas, J. A., Palomares, N. A., Richards, A. S., Keating, D. M., Joyce,
with a representative sample, which will help identify dimen- N., & Rains, S. A. (2022). When machine and bandwagon heuristics
sions on which people evaluate machines, either similarly or compete: Understanding users’ response to conflicting AI and crowd-
differently, vis-à-vis humans. In this regard, Hong et al. sourced fact-checking. Human Communication Research, 48(3),
(2022) proposed “creative machine heuristics” to capture 430–461. https://doi.org/10.1093/hcr/hqac010
what human-like traits people expect of machines, focusing Banks, J., Edwards, A. P., & Westerman, D. (2021). The space between:
Nature and machine heuristics in evaluations of organisms, cyborgs,
on creativity in particular. Although the attempt to expand and robots. Cyberpsychology, Behavior, and Social Networking,
the repertoire of machine heuristic seems well justified, they 24(5), 324–331. https://doi.org/10.1089/cyber.2020.0165
focused on “how people think about machines being crea- Bellur, S., & Sundar, S. S. (2014). How can we tell when a heuristic has

Downloaded from https://academic.oup.com/hcr/article/50/2/184/7329158 by guest on 26 January 2025

tive” (p. 2) rather than how creative people think machines been used? Design and analysis strategies for capturing the operation
are (or can be). of heuristics. Communication Methods and Measures, 8(2),
If machine heuristic serves as a moderator of the source 116–137. https://doi.org/10.1080/19312458.2014.903390
effects, something one brings into the communication context, Berger, C. R., Roloff, M. E., & Roskos-Ewoldsen, D. R. (2010). What is
communication science? In C. R. Berger, M. E. Roloff, & D. R.
then what would contribute to such beliefs? Usual suspects in-
Roskos-Ewoldsen (Eds.), The handbook of communication science
clude the exposure to media discourse, education, computer (pp. 3–20). Sage.
proficiency, and direct or indirect experiences with machines. Brandtzaeg, P. B., Skjuve, M., &Følstad, A. (2022). My AI friend: How
Less obvious is the role of dissatisfaction with a human equiv- users of a social chatbot understand their human–AI friendship.
alent. In one study (Lee et al., 2022), for instance, after wit- Human Communication Research, 48(3), 404–429. https://doi.org/
nessing a human moderator’s unsatisfactory performance, 10.1093/hcr/hqac008
Chaiken, S. (1980). Heuristic versus systematic information processing
participants were more likely to endorse AI heuristic, suggest-
and the use of source versus message cues in persuasion. Journal of
ing that people consider AI as a potential replacement of sub- Personality and Social Psychology, 39(5), 752–766. https://doi.org/
optimal human agents. Possibly, disappointment with human 10.1037/0022-3514.39.5.752
performance may lead people to project more optimistic or Chaiken, S. (1987). The heuristic model of persuasion. In M. P. Zanna,
wishful images unto AI. J. M. Olson, & C. P. Herman (Eds.), Social influence: The Ontario
In celebration of Human Communication Research’s 50th symposium (Vol. 5, pp. 3–39). Lawrence Erlbaum Associates.
anniversary, this special issue is dedicated to innovative theory Chaiken, S., Lieberman, A., & Eagly, A. H. (1989). Heuristic and sys-
tematic information processing within and beyond the persuasion
development. Since the pioneering book by Reeves and Nass
context. In J. S. Uleman & J. A. Bargh (Eds.), Unintended thought
(1996) was published, numerous studies have tested the CASA (pp. 212–252). Guilford Press.
paradigm in a wide range of contexts, lending support to its Cialdini, R. B. (2007). Influence: The psychology of persuasion. Collins.
key proposition. Still, replicating a particular phenomenon (in Cloudy, J., Banks, J., & Bowman, N. D. (2021). The str(AI)ght scoop:
this case, social responses to computers) across contexts, albeit Artificial intelligence cues reduce perceptions of hostile media bias.
attesting to its regularity, falls short of explaining why the ob- Digital Journalism. https://doi.org/10.1080/21670811.2021.
served phenomenon occurs. If “identified regularities need to 1969974
DeVito, M. A., Gergle, D., & Birnholtz, J. (2017). “Algorithms ruin
be explained by recourse to mechanisms that account for the
everything”: #RIPTwitter, folk theories, and resistance to algorith-
regularity in question” (Berger et al., 2010, p. 10) for theory mic change in social media. In Proceedings of the 2017 CHI
construction, then the psychological processes proposed herein Conference on Human Factors in Computing Systems, USA (pp.
may serve as one such mechanism, if validated through rigor- 3163–3174). https://doi.org/10.1145/3025453.3025659
ous empirical testing. In this sense, the current essay represents Fiske, S. T., & Taylor, S. E. (1984). Social cognition. McGraw-Hill.
a long over-due attempt to update the CASA framework that Fogg, B. J., & Nass, C. (1997a). How users reciprocate to computers:
paved the way for HMC research almost three decades ago. An experiment that demonstrates behavior change. In CHI ’97 ex-
tended abstracts on human factors in computing systems (pp.
Not only does the extent to which people adjust their responses
331–332), USA. https://doi.org/10.1145/1120212.1120419
to machines in consideration of their nonhuman nature vary as Fogg, B. J., & Nass, C. (1997b). Silicon sycophants: The effects of com-
a function of mindlessness, but the degree to which people con- puters that flatter. International Journal of Human–Computer
sider the nonhumanness as a relevant factor in their communi- Studies, 46(5), 551–561. https://doi.org/10.1006/ijhc.1996.0104
cation may also vary, depending on how mindful they are (or Hancock, J. T., Naaman, M., & Levy, K. (2020). AI-mediated commu-
choose to be). nication: Definition, research agenda, and ethical considerations.
Journal of Computer-Mediated Communication, 25(1), 89–100.
https://doi.org/10.1093/jcmc/zmz022
Higgins, E. T. (1996). Knowledge activation: Accessibility, applicability,
Funding and salience. In E. T. Higgins & A. W. Kruglanski (Eds.), Social psy-
chology: Handbook of basic principles (pp. 133–168). Guilford
This work was supported by the National Research Press.
Foundation of Korea (NRF) grant funded by the Korea gov- Hilton, J. L., Fein, S., & Miller, D. T. (1993). Suspicion and disposi-
ernment (MSIT) (No. 2022R1A5A708390812). tional inference. Personality and Social Psychology Bulletin, 19(5),
501–512. https://doi.org/10.1177/0146167293195003
Conflicts of interest: No potential conflict of interest was Hong, J.-W., Fischer, K., Ha, Y., & Zeng, Y. (2022). Human, I wrote a
reported by the author. song for you: An experiment testing the influence of machines’
192 Minding the source in HMC

attributes on the AI-composed music evaluation. Computers in Journal of Pragmatics, 59(Part B), 210–220. https://doi.org/10.
Human Behavior, 131, 107239. https://doi.org/10.1016/j.chb.2022. 1016/j.pragma.2013.07.012
107239 Molina, M. D., & Sundar, S. S. (2022). Does distrust in humans predict
Huang, S. A., Hancock, J., & Tong, S. T. (2022). Folk theories of online greater trust in AI? Role of individual differences in user responses to
dating: Exploring people’s beliefs about the online dating process content moderation. New Media & Society. https://doi.org/10.1177/
and online dating algorithms. Social Media þ Society, 8(2), 1–12. 14614448221103534
https://doi.org/10.1177/20563051221089561 Moon, Y., & Nass, C. (1996). How “Real” are computer personalities?
Johnson, D., Gardner, J., & Wiles, J. (2004). Experience as a moderator Psychological responses to personality types in human–computer in-
of the media equation: The impact of flattery and praise. teraction. Communication Research, 23(6), 651–674. https://doi.
International Journal of Human–Computer Studies, 61(3), org/10.1177/009365096023006002
237–258. https://doi.org/10.1016/j.ijhcs.2003.12.008 Morkes, J., Kernal, H. K., & Nass, C. (1999). Effects of humor in
Jones-Jang, S. M., & Park, Y. J. (2023). How do people react to AI fail- task-oriented human–computer interaction and computer-mediated
ure? Automation bias, algorithmic aversion, and perceived controlla- communication: A direct test of SRCT theory. Human–Computer
bility. Journal of Computer-Mediated Communication, 28(1), 1–8. Interaction, 14(4), 395–435. https://doi.org/10.1207/S15327051H
https://doi.org/10.1093/jcmc/zmac029 CI1404_2
Kruglanski, A. W., & Thompson, E. P. (1999). Persuasion by a single Nass, C., Moon, Y., & Green, N. (1997). Are machines gender neutral?

Downloaded from https://academic.oup.com/hcr/article/50/2/184/7329158 by guest on 26 January 2025

route: A view from the unimodel. Psychological Inquiry, 10(2), Gender-stereotypic responses to computers with voices. Journal of
83–109. https://doi.org/10.1207/s15327965pl100201 Applied Social Psychology, 27(10), 864–876. https://doi.org/10.
Langer, E. J., Blank, A., & Chanowitz, B. (1978). The mindlessness of 1111/j.1559-1816.1997.tb00275.x
ostensibly thoughtful action: The role of “placebic” information in Nass, C., Moon, Y., & Carney, P. (1999). Are people polite to com-
interpersonal interaction. Journal of Personality and Social puters? Responses to computer-based interviewing systems. Journal
Psychology, 36(6), 635–642. https://doi.org/10.1037/0022-3514.36. of Applied Social Psychology, 29(5), 1093–1109. https://doi.org/10.
6.635 1111/j.1559-1816.1999.tb00142.x
Lee, E.-J. (2008). Gender stereotyping of computers: Resource depletion Nass, C., & Moon, Y. (2000). Machines and mindlessness: Social
or reduced attention? Journal of Communication, 58(2), 301–320. responses to computers. Journal of Social Issues, 56(1), 81–103.
https://doi.org/10.1111/j.1460-2466.2008.00386.x https://doi.org/10.1111/0022-4537.00153
Lee, E.-J. (2010). What triggers social responses to flattering computers? Petty, R. E., & Cacioppo, J. T. (1986). Communication and persuasion:
Experimental tests of anthropomorphism and mindlessness explana- Central and peripheral routes to attitude change. Springer.
tions. Communication Research, 37(2), 191–214. https://doi.org/10. Petty, R. E., Cacioppo, J. T., & Goldman, R. (1981). Personal involve-
1177/0093650209356389 ment as a determinant of argument-based persuasion. Journal of
Lee, E.-J. (2020). Authenticity model of (mass-oriented) Personality and Social Psychology, 41(5), 847–855. https://doi.org/
computer-mediated communication: Conceptual explorations 10.1037/0022-3514.41.5.847
and testable propositions. Journal of Computer-Mediated Petty, R. E., Wheeler, S. C., & Bizer, G. Y. (1999). Is there one persua-
Communication, 25(1), 60–73. https://doi.org/10.1093/jcmc/ sion process or more? Lumping versus splitting in attitude change
zmz025 theories. Psychological Inquiry, 10(2), 156–163. https://doi/org/10.
Lee, E.-J., Kim, H. S., & Joo, M. H. (2023a). Social media vs. mass me- 1207/S15327965PL100211
dia: Mitigating the suspicion of ulterior motives in public health Reeves, B., & Nass, C. (1996). The media equation: How people treat
communication. Health Communication, 38(11), 2450–2460. computers, television, and new media like real people and places.
https://doi.org/10.1080/10410236.2022.2074781 Cambridge University Press.
Lee, E.-J., Kim, H. S., Suh, Y. J., & Park, J. W. (2022). Something’s fishy Sundar, S. S. (2008). The MAIN model: A heuristic approach to under-
about it: How opinion congeniality and explainability affect moti- standing technology effects on credibility. In M. J. Metzger & A. J.
vated attribution to artificial intelligence versus human comment Flanagin (Eds.), Digital media, youth, and credibility (pp. 72–100).
moderators. Cyberpsychology, Behavior, and Social Networking, MIT Press.
25(8), 496–503. https://doi.org/10.1089/cyber.2021.0347 Sundar, S. S. (2020). Rise of machine agency: A framework for studying
Lee, E.-J., Kim, H. S., Suh, Y. J., & Park, J. (2023b, May 25–29). Not so the psychology of human–AI interaction (HAII). Journal of
automatic perception of hostility in automated journalism: The role Computer-Mediated Communication, 25(1), 74–88. https://doi.org/
of source congeniality and belief in AI heuristic. In The 73rd annual 10.1093/jcmc/zmz026
conference of international communication association, Toronto, Sundar, S. S., & Kim, J. (2019). Machine heuristic: When we trust com-
Canada. puters more than humans with our personal information. In
Lee, E.-J., & Nass, C. (2002). Experimental tests of normative group in- Proceedings of the 2019 CHI conference on human factors in com-
fluence and representation effects in computer-mediated communica- puting systems (pp. 1–9), UK. https://doi.org/10.1145/3290605.
tion: When interacting via computers differs from interacting with 3300768
computers. Human Communication Research, 28(3), 349–381. Sundar, S. S., & Lee, E.-J. (2022). Rethinking communication in the era
https://doi.org/10.1111/j.1468-2958.2002.tb00812.x of artificial intelligence. Human Communication Research, 48(3),
Lee, S., Oh, J., & Moon, W.-K. (2023). Adopting voice assistants in on- 379–385. https://doi.org/10.1093/hcr/hqac014
line shopping: Examining the role of social presence, performance Sundar, S. S., & Nass, C. (2000). Source orientation in human–com-
risk, and machine heuristic. International Journal of puter interaction: Programmer, networker, or independent social ac-
Human–Computer Interaction, 39(14), 2978–2992. https://doi.org/ tor. Communication Research, 27(6), 683–703. https://doi.org/10.
10.1080/10447318.2022.2089813 1177/009365000027006001
Liu, B., & Wei, L. (2018). Reading machine-written news: Effect of ma- Sundar, S. S., Xu, Q., & Dou, X. (2019). Role of technology in online
chine heuristic and novelty on hostile media perception. In M. persuasion: A MAIN model perspective. In S. Rodgers & E. Thorson
Kurosu (Ed.), Human–computer interaction. Theories, methods, and (Eds.), Advertising theory (2nd ed., pp. 70–88). Routledge.
human issues (pp. 307–324). Springer. Tandoc, E. C., Jr., Yao, L. J., & Wu, S. (2020). Man vs. machine? The
Metzger, M. J., Flanagin, A. J., & Medders, R. B. (2010). Social and impact of algorithm authorship on news credibility. Digital
heuristic approaches to credibility evaluation online. Journal of Journalism, 8(4), 548–562. https://doi.org/10.1080/21670811.20
Communication, 60(3), 413–439. https://doi.org/10.1111/j.1460- 20.1762102
2466.2010.01488.x Trafimow, D. (2013). The importance of surprising findings in psychol-
Metzger, M. J., & Flanagin, A. J. (2013). Credibility and trust of infor- ogy. Frontiers in Psychology, 4(173), 1–2. https://doi.org/10.3389/
mation in online environments: The use of cognitive heuristics. fpsyg.2013.00173
Human Communication Research, (2024), Vol. 50, No. 2 193

Tversky, A., & Kahenman, D. (1973). Availability: A heuristic for judging and credibility in news content. Digital Journalism, 9(1), 64–83.
frequency and probability. Cognitive Psychology, 5(2), 207–232. https://doi.org/10.1080/21670811.2020.1851279
https://doi.org/10.1016/0010-0285(73)90033-9 Weise, K., & Metz, C. (2023, May 9). When A.I. Chatbots hallucinate.
Turing, A. M. (1950). I—Computing machinery and intelligence. Mind, The New York Times. https://www.nytimes.com/2023/05/01/busi
LIX(236), 433–460. https://doi.org/10.1093/mind/LIX.236.433 ness/ai-chatbots-hallucination.html
Waddell, T. F. (2018). A robot wrote this? Digital Journalism, 6(2), Xu, K., Chen, X., & Huang, L. (2022). Deep mind in social responses to
236–255. https://doi.org/10.1080/21670811.2017.1384319 technologies: A new approach to explaining the computers are social
Wang, S. (2021). Moderating uncivil user comments by humans or actors phenomena. Computers in Human Behavior, 134, 1–13.
machines? The effects of moderation agent on perceptions of bias https://doi.org/10.1016/j.chb.2022.107321

Downloaded from https://academic.oup.com/hcr/article/50/2/184/7329158 by guest on 26 January 2025