Computers in Human Behavior: Artificial Humans最新文献

{"title":"Harnessing ChatGPT for workplace innovation: The mediating role of internal work locus of control","authors":"Hwajin Yang,&nbsp;Zhi Yi Koh,&nbsp;Frosch Y.X. Quek","doi":"10.1016/j.chbah.2026.100373","DOIUrl":"10.1016/j.chbah.2026.100373","url":null,"abstract":"<div><div>Given the rapid integration of artificial intelligence (AI) technologies in the workplace, this study examined associations between ChatGPT use and innovative work behavior, with particular attention to internal work locus of control as a potential psychological pathway underlying this relationship. Using a cross-sectional, single-wave survey of 179 employees recruited in Singapore, we examined ChatGPT-use frequency and perceived effectiveness, internal work locus of control, and three facets of innovative work behavior—idea generation, idea promotion, and idea implementation. Structural equation models showed that both ChatGPT-use indicators were positively associated with internal work locus of control. In turn, internal work locus of control was positively associated with idea promotion and implementation, resulting in significant indirect associations for these two stages but not for idea generation. These results remained robust after controlling for key covariates (age, sex, subjective socioeconomic status, employment tenure, work amotivation, and burnout). These findings suggest that the association between ChatGPT use and workplace innovation may differ across innovation stages and highlight internal work locus of control as a potentially relevant psychological correlate of idea promotion and implementation.</div></div>","PeriodicalId":100324,"journal":{"name":"Computers in Human Behavior: Artificial Humans","volume":"9 ","pages":"Article 100373"},"PeriodicalIF":0.0,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148935065","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Loneliness and generative AI use for companionship among college students","authors":"Cindy H. Liu,&nbsp;Chang Zhao,&nbsp;Felix Lou,&nbsp;Wenbo Zhang,&nbsp;Anthony Menor,&nbsp;Angela Chow,&nbsp;Tiffany Yip","doi":"10.1016/j.chbah.2026.100348","DOIUrl":"10.1016/j.chbah.2026.100348","url":null,"abstract":"<div><div>Generative AI chatbots are a new class of artificial agents that young adults may turn to for emotional companionship, raising questions for the study of interaction between humans and AI. Whether loneliness uniquely predicts AI use for companionship, and whether any such association is specific to companionship-seeking rather than general AI use, has not been examined in a U.S. college sample. Using data from the Healthy Minds Study (HMS) 2024-2025, we examined predictors of AI companionship use among U.S. college students (N = 858). Hierarchical logistic regression tested whether loneliness was uniquely associated with AI companionship use after adjusting for demographics and mental health symptoms, and whether the same association held for general AI use. Seventeen percent of students reported using AI for companionship. Loneliness was uniquely associated with AI companionship use, but not with general AI use, after full adjustment (OR = 1.21, 95% CI [1.08, 1.36], p &lt; .01), indicating specificity to companionship-seeking. Neither campus belonging nor satisfaction with social relationships predicted such use. Asian and international students were more likely to use AI for companionship, whereas LGBQ+ students were less likely. In a combined model directly contrasting non-users, general-only users, and companionship users, loneliness distinguished companionship users specifically (p &lt; 0.01) and predicted more frequent companionship use in a dose-dependent pattern. These findings position AI companionship as a distinct form of human-AI interaction. Because the data are cross-sectional, we frame loneliness as a candidate predictor to be confirmed in longitudinal research.</div></div>","PeriodicalId":100324,"journal":{"name":"Computers in Human Behavior: Artificial Humans","volume":"9 ","pages":"Article 100348"},"PeriodicalIF":0.0,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148636590","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Comparable costly punishment toward AI and human free-riders among Japanese participants: Evidence from a public goods game","authors":"Koshiro Tanaka,&nbsp;Takamasa Iio,&nbsp;Masahiro Shiomi","doi":"10.1016/j.chbah.2026.100361","DOIUrl":"10.1016/j.chbah.2026.100361","url":null,"abstract":"<div><div>Costly punishment, in which individuals incur a personal cost to sanction norm violators, is a key mechanism for maintaining cooperation in social dilemmas. Although such behavior has been extensively studied in human–human interaction, it remains unclear whether people direct comparable costly punishment toward artificial intelligence (AI), particularly within culturally specific contexts. The present study examined this question among Japanese participants.</div><div>We conducted an online public goods game involving 588 Japanese participants who interacted with three pre-programmed agents presented as either humans or AI. Participants could incur a personal cost to report a free-rider, and this reporting behavior was used as an observable measure of costly punishment. A three-way ANOVA, equivalence testing (TOST), regression analyses, and supplementary time-series analyses were performed.</div><div>The results showed that costly punishment toward AI free-riders was statistically and practically equivalent to that directed toward human free-riders. Participants’ behavior was primarily influenced by whether another agent had already punished the free-rider, regardless of whether that agent was presented as human or AI. Supplementary analyses further showed that this behavioral equivalence was already evident during the early rounds, while time-series analyses indicated that participants' behavioral patterns stabilized after the initial phase of repeated interaction, suggesting that the equivalence was not limited to either initial impressions or later repeated interactions. In contrast, questionnaire responses revealed differences in some subjective evaluations despite comparable behavioral responses.</div><div>These findings suggest that, among Japanese participants under tightly controlled label-based conditions, simply presenting an interaction partner as AI rather than human did not substantially alter observable costly punishment. The study extends research on negative human–AI interaction by showing that AI can function as a socially accountable target of costly punishment in a Japanese sample, while emphasizing that the generalizability of this equivalence to other cultural contexts requires cross-cultural validation.</div></div>","PeriodicalId":100324,"journal":{"name":"Computers in Human Behavior: Artificial Humans","volume":"9 ","pages":"Article 100361"},"PeriodicalIF":0.0,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148636655","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"How workplace AI feedback shapes AI dependency: A randomized controlled trial on the moderating role of workplace social support among Japanese office workers","authors":"Yasushi Watanabe,&nbsp;Masataka Nakayama,&nbsp;Kosuke Takemura,&nbsp;Yukiko Uchida","doi":"10.1016/j.chbah.2026.100365","DOIUrl":"10.1016/j.chbah.2026.100365","url":null,"abstract":"<div><div>AI dependency—habitual, automatic AI use accompanied by diminished autonomy—is a growing concern in the workplace, yet the conditions under which it develops remain poorly understood. Two competing pathways have been proposed. Users with limited social support from colleagues may compensate for this deficit by depending on AI (the <em>compensatory pathway</em>), or users with abundant support may be better positioned to leverage AI interaction, thereby fostering dependency (the <em>enhancement pathway</em>). To test these pathways, we analyzed data from a 5-day randomized controlled trial with 294 Japanese office workers. Participants engaged in daily work-reflection sessions with an AI chatbot. Depending on condition, the chatbot delivered positive feedback emphasizing strengths, negative feedback highlighting areas for improvement, or no work-related feedback (control). AI dependency was measured one week after the final session, revealing two qualitatively distinct patterns based on feedback valence. Positive feedback increased AI dependency regardless of workplace social support levels, suggesting that positive reinforcement from AI functions as a general pathway to dependency. In contrast, negative feedback did not increase AI dependency overall but did so selectively among participants with abundant social support, which is consistent with the enhancement pathway. These findings are qualified by the short-term, self-reported outcome, the exploratory and non-preregistered analysis, the modest size of the focal interaction, and the crowdsourced Japanese sample, which limits their generalizability. Within these bounds, they indicate that AI dependency is not simply a function of AI characteristics but emerges from the interaction between AI feedback and employees' existing workplace relationships.</div></div>","PeriodicalId":100324,"journal":{"name":"Computers in Human Behavior: Artificial Humans","volume":"9 ","pages":"Article 100365"},"PeriodicalIF":0.0,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148738634","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"How prompting and editing shape psychological ownership in AI-assisted creative writing","authors":"Niki Pennanen,&nbsp;Pyry Kanerva,&nbsp;Christian Guckelsberger","doi":"10.1016/j.chbah.2026.100381","DOIUrl":"10.1016/j.chbah.2026.100381","url":null,"abstract":"<div><div>Large language models are increasingly used as writing partners, but their use complicates whether writers experience the resulting text as genuinely their own. Such psychological ownership is especially relevant in AI-assisted writing because it contributes to the immediate and extended experience of AI use, and users may sign, share, revise, or be held accountable for texts whose content was partly generated by AI. Prior work shows that greater human input can increase ownership, but it remains unclear how ownership is shaped by two common forms of involvement: steering generation through prompting, directly revising generated text, and their combination. We report a preregistered mixed factorial experiment (<span><math><mrow><mi>N</mi><mo>=</mo><mn>273</mn></mrow></math></span>) in which online participants wrote short stories either without AI assistance or with AI assistance under one of two prompting styles (single-prompt vs. iterative), crossed with editing or not editing the AI-generated text. Editing increased ownership, especially for single-prompt generation. Iterative prompting increased ownership when editing was unavailable, but not when editing was available. These findings suggest that iterative prompting and post-generation editing are partially substitutable routes to psychological ownership, each providing opportunities for control and self-investment at different phases of the writing process. Fully manual writing nevertheless produced substantially higher ownership than all AI-assisted conditions, a gap that persisted in exploratory covariate analyses adjusting for effort, perceived story quality, and intrinsic motivation. The results suggest that AI writing tools should avoid one-shot delegation workflows when ownership matters, and instead support users in both steering and directly revising co-created text.</div></div>","PeriodicalId":100324,"journal":{"name":"Computers in Human Behavior: Artificial Humans","volume":"9 ","pages":"Article 100381"},"PeriodicalIF":0.0,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148843743","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"From optimization to explanation: A theory-oriented human–agent dialogue (T-HAD) framework in the era of large language models","authors":"Tomoki Miyamoto","doi":"10.1016/j.chbah.2026.100375","DOIUrl":"10.1016/j.chbah.2026.100375","url":null,"abstract":"<div><div>The rapid proliferation of large language models (LLMs) has transformed human–agent dialogue (HAD) research, enabling agents to reflect the complexity of human social interaction. Traditionally, the field has been driven by a “practice-oriented” paradigm (P-HAD) focused on engineering optimization to enhance task efficiency and user satisfaction. While numerous studies have individually explored “theory-oriented” aspects (T-HAD), the epistemological distinction between these two approaches remains ambiguous. This challenge is further intensified by the opacity of LLM-based systems. This study proposes a systematic conceptual and methodological framework that conceptually distinguishes explanation-oriented (T-HAD) and optimization-oriented (P-HAD) research according to the role of theory and the primary research objective, using the analogy of “prediction” versus “explanation” as a heuristic rather than a strict epistemological mapping. To facilitate the practical application of this framework, operational decision criteria are introduced for distinguishing T-HAD from P-HAD. Using these criteria, a targeted bibliometric analysis of dialogue-focused papers published in the Human-Agent Interaction conference proceedings from 2020 to 2025 indicates that T-HAD studies constitute a minority of the examined dialogue literature. Using politeness theory as a case study, this study illustrates how the same theory can serve as an engineering heuristic in P-HAD while providing empirically examinable explanatory propositions in T-HAD. A T-HAD research protocol is presented to guide the construction, empirical examination, and refinement of theoretical accounts of HAD. Finally, this study introduces theory-driven dialogue prompts to support systematic experimental manipulation in LLM-based T-HAD research and the translation of explanatory knowledge into dialogue-agent design. These contributions provide conceptual and methodological guidance for advancing T-HAD research in the LLM era and may contribute to more interpretable and responsible dialogue-agent development.</div></div>","PeriodicalId":100324,"journal":{"name":"Computers in Human Behavior: Artificial Humans","volume":"9 ","pages":"Article 100375"},"PeriodicalIF":0.0,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148934760","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"“Ban it!”: Moral foundations of preemptive objections to emerging warbot technology","authors":"Stephanie M. Merritt,&nbsp;Nathan L. Tenhundfeld","doi":"10.1016/j.chbah.2026.100374","DOIUrl":"10.1016/j.chbah.2026.100374","url":null,"abstract":"<div><div>As technology advances, questions about whether machines can and should make life-and-death decisions are gaining increasing attention. This investigation focuses on autonomous military technologies (warbots), which are in development but not yet broadly deployed. The very concept of warbots has sparked debate and protest, with tens of thousands signing petitions for preemptive bans on their development. Understanding preemptive resistance to emerging technologies is an important yet understudied area of inquiry. As such, the goal of this research is to illuminate sources of preemptive attempts to ban development, using warbots as the specific focus. Drawing upon Moral Foundations Theory and the five foundations of morality (care/harm, fairness/cheating, loyalty/betrayal, authority/subversion, and sanctity/degradation), we examine which of the foundations are most relevant to preemptive warbot objections in a sample of N = 289 US and Canadian adults. We find that the sanctity/degradation foundation is key, such that perceived violations of the sanctity foundation accounted for unique variance in moral wrongness judgments over and above the other four foundations. We also sought to understand what people imagine when they think of warbots, as their imaginings likely influence perceived moral violations. Importantly, most participants reported imagining warbots that were humanoid in form more than they reported imagining other forms, and they felt humanoid warbots were creepier than other forms. Our findings suggest preemptive objections are influenced not only by intuitions that warbots are harmful but also by intuitions that they are degrading and creepy.</div></div>","PeriodicalId":100324,"journal":{"name":"Computers in Human Behavior: Artificial Humans","volume":"9 ","pages":"Article 100374"},"PeriodicalIF":0.0,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148934815","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Erratum regarding corrections related to Ethical approval and informed consent in articles published in [Computers in Human Behavior: Artificial Humans 6C (2025)]","authors":"","doi":"10.1016/j.chbah.2026.100354","DOIUrl":"10.1016/j.chbah.2026.100354","url":null,"abstract":"","PeriodicalId":100324,"journal":{"name":"Computers in Human Behavior: Artificial Humans","volume":"9 ","pages":"Article 100354"},"PeriodicalIF":0.0,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148934820","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Distinct patterns of social media use in later life: Digitally mediated social environments and a dual-association pattern of human vulnerability","authors":"Yuna Seo,&nbsp;Yuki Nakada","doi":"10.1016/j.chbah.2026.100313","DOIUrl":"10.1016/j.chbah.2026.100313","url":null,"abstract":"<div><div>As digitally mediated social environments increasingly structure everyday interaction, understanding how such artificial environments relate to human vulnerability has become a critical challenge. This study examines how distinct patterns of social media engagement, active/expressive versus informational-communicative use, are associated with frailty among older adults, and whether psychological well-being and lifestyle behaviors operate as intervening associations rather than causal pathways. Survey data were collected from 963 community-dwelling Japanese adults aged 65–89. Exploratory factor analysis identified two dimensions of social media use. Logistic regression using a quartile-based frailty definition showed that informational-communicative social media use was associated with a higher likelihood of frailty, whereas active/expressive use was not. However, this direct association was not robust in sensitivity analyses using the established clinical Kihon Checklist cutoff; under this clinical criterion, informational-communicative use was no longer significantly associated with frailty and the odds ratio changed direction. In contrast, psychological well-being and healthy lifestyle behaviors were consistently associated with lower frailty across operationalizations. Mediation analyses conducted within a structural equation modeling framework suggested a modest indirect protective association through psychological well-being and healthier lifestyle behaviors. Taken together, the findings indicate a definition-sensitive and dual-association pattern, rather than a robust direct association between informational-communicative use and clinically defined frailty. Cluster analysis further demonstrated substantial heterogeneity in digital engagement, with high-use individuals exhibiting the highest frailty prevalence under the quartile-based frailty classification. These findings should be interpreted as a pattern of associations rather than causal pathways, as the cross-sectional design does not permit temporal inference; alternative explanations, including reverse causality, remain plausible. Nevertheless, the results suggest that artificially mediated social environments may be linked to heterogeneous patterns of human vulnerability in later life, particularly at the level of relative or subclinical vulnerability rather than clinically defined frailty. This nuanced association pattern underscores the importance of age-sensitive design and governance of digital social systems.</div></div>","PeriodicalId":100324,"journal":{"name":"Computers in Human Behavior: Artificial Humans","volume":"8 ","pages":"Article 100313"},"PeriodicalIF":0.0,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147802893","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Can we rely on generative AI for emergency patient triage classification?","authors":"Udochukwu Nwandikom,&nbsp;Akbar Siami Namin","doi":"10.1016/j.chbah.2026.100301","DOIUrl":"10.1016/j.chbah.2026.100301","url":null,"abstract":"<div><div>Triage involves classifying patients based on their need for treatment and the available medical resources. Owing to the rampant issue of Emergency Department (ED) overcrowding, medical sciences pursue faster and more consistent approaches to patient triage. Advances in generative artificial intelligence, particularly large language models (LLMs), have prompted interest in their potential use for clinical decision support. However, the <em>humanness</em> of these systems can foster human-like trust, which may overlook reliability and functional performance in favor of perceived competence and integrity. This study evaluates whether such trust is justified by examining the use of LLMs for emergency triage prediction. We compare LLM and expert triage decisions across 111 mass-casualty scenarios using the START/JUMPSTART and National Triage Scale (NTS) systems. Agreement between model predictions and expert solutions was evaluated using the Quadratic Weighted Kappa (QWK) statistic with expert-derived weight penalties. The results show moderate agreement between LLM predictions and expert triage decisions (QWK = 0.44), indicating that the tested model may not be sufficiently reliable for autonomous triage classification. An explainability analysis of the model responses suggests that while the model performs adequately in controlled scenarios, it struggles in typical mass-casualty situations, often prioritizing textual cues over triage assignment logic. These findings support calibrated trust in LLM-assisted triage and the continued need for human oversight. The flaws we have observed in the model reasoning pathways for triage decision-making make an argument for neuro-symbolic machine learning.</div></div>","PeriodicalId":100324,"journal":{"name":"Computers in Human Behavior: Artificial Humans","volume":"8 ","pages":"Article 100301"},"PeriodicalIF":0.0,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148178287","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}