Computers in human behavior reports最新文献

{"title":"Can AI reduce gender bias? The case of AI-generated audio conversations in Google's NotebookLM","authors":"Milad Malekzadeh","doi":"10.1016/j.chbr.2026.101246","DOIUrl":"10.1016/j.chbr.2026.101246","url":null,"abstract":"<div><div>Voice-based generative AI increasingly mediates how people encounter scientific knowledge, but its conversational designs may reproduce gendered patterns of authority even when content appears neutral. This study audits Google NotebookLM's Audio Overview feature to examine whether AI-generated podcast-style conversations distribute speaking time, epistemic authority, and communicative roles equitably across woman-coded and man-coded voices. 100 default audio summaries were generated from anonymized scientific papers, then transcribed, diarized, and coded the conversations for speaking time, technical level, speech acts, utterance-level epistemic positioning, and conversation-level roles. To contextualize the findings, the same workflow was applied to 100 episodes from highly ranked human-produced science podcasts. NotebookLM exhibited near parity in airtime: woman-coded voices accounted for 48.9% of speaking time and man-coded voices for 51.1%, with no statistically reliable difference after correction for multiple comparisons. By contrast, in the human podcast reference corpus, woman-coded speakers accounted for 38.4% of gender-coded speaking time and man-coded speakers for 61.6%, and this imbalance remained statistically detectable after correction. For the principal comparisons concerning self-positioning as expert, conversation-level epistemic authority, communicative roles, technical level, question proportion, and speech-act distributions, the corrected analyses did not provide statistically reliable evidence of gender-coded differences. These findings indicate that the clearest contrast between the audited NotebookLM outputs and the human podcast reference corpus concerned airtime: NotebookLM showed substantially greater gender-coded speaking-time parity, while evidence of broader interactional asymmetry was limited.</div></div>","PeriodicalId":72681,"journal":{"name":"Computers in human behavior reports","volume":"23 ","pages":"Article 101246"},"PeriodicalIF":5.5,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148736328","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":"What shapes what we know about emoji communication? A structural topic model of thematic evolution and institutional forces","authors":"Pengyu Chu,&nbsp;Yuning Sun","doi":"10.1016/j.chbr.2026.101235","DOIUrl":"10.1016/j.chbr.2026.101235","url":null,"abstract":"<div><div>Emoji are a central resource for nonverbal expression in digital communication, yet the research landscape remains fragmented and has never been computationally mapped. This study applies structural topic modeling to 937 emoji articles indexed in Web of Science, identifying latent topics and modeling the effects of publication year, region, funding, and collaboration type on topic prevalence. Twelve distinct topics emerged spanning communicative functions, cognitive processing, social media discourse, and health communication. Covariate analyses revealed that institutional forces systematically shape the field: marketing and education topics are rising while food and sensory science is declining; funded research is associated with applied rather than humanistic topics; and single-authored work concentrates on theoretical analysis. We propose the Encoding–Decoding–Impact framework, organizing the literature along the communicative arc from cue production through interpretation to downstream effects. This framework exposes how the field’s trajectory is shaped not only by intellectual priorities but by the political economy of knowledge production.</div></div>","PeriodicalId":72681,"journal":{"name":"Computers in human behavior reports","volume":"23 ","pages":"Article 101235"},"PeriodicalIF":5.5,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148736338","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":"Beyond awareness: A Bayesian cross-classified analysis of contextual factors and preservice teachers’ AI tool familiarity","authors":"Moeketsi Mosia,&nbsp;Fadip Audu Nannim,&nbsp;Felix Egara,&nbsp;Moses Basitere","doi":"10.1016/j.chbr.2026.101262","DOIUrl":"10.1016/j.chbr.2026.101262","url":null,"abstract":"<div><div>Preservice mathematics teachers may recognise the educational affordances of artificial intelligence (AI) while remaining unevenly familiar with specific AI tools. This exploratory cross-sectional study examined how awareness, mentor support, structural pressure, self-initiative, prior training, and informal AI use relate to self-reported familiarity across five mathematics-oriented AI tool categories. We fitted a Bayesian cross-classified cumulative ordinal model to 650 person-tool observations from 130 South African preservice mathematics teachers, treating familiarity as jointly structured by persons and tools. Awareness showed the strongest positive association with familiarity, with an odds ratio (OR) of 2.41 and a 95% credible interval (CrI) of 1.50 to 3.94. Structural pressure was also positively associated with familiarity (β = 0.67, 95% CrI [0.12, 1.16]), but this pattern is interpreted cautiously because reverse causality, omitted variables, and construct ambiguity remain plausible. Tool-level heterogeneity was substantial: problem-solving apps had consistently higher probabilities of reported familiarity (about 44–63% across scenarios) than symbolic engines and explanation bots (below 14%). The findings are therefore presented as hypothesis-generating evidence for tool-specific teacher education and for future longitudinal, experimental, and performance-based research, not as causal guidance for programme design.</div></div>","PeriodicalId":72681,"journal":{"name":"Computers in human behavior reports","volume":"23 ","pages":"Article 101262"},"PeriodicalIF":5.5,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148776484","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":"Problematic video-streaming in adolescents: The role of binge-watching patterns, short video format, and emotion regulation abilities","authors":"Jan-Ole Cloes,&nbsp;Lisa Klamert,&nbsp;Kerstin Paschke","doi":"10.1016/j.chbr.2026.101247","DOIUrl":"10.1016/j.chbr.2026.101247","url":null,"abstract":"<div><h3>Background</h3><div>Viewing videos online, i.e., video streaming, is a popular internet activity among adolescents. However, this age group is especially vulnerable to developing problematic internet use (PIU). Previous research highlights the relevance of emotion regulation (ER) across different forms of PIU but much less is known about problematic video streaming (PVS). Excessive binge-watching (BW) is linked to deficient ER and addiction in adults. Yet, the diversification of online video formats leads to new concerns and the need for extended research.</div></div><div><h3>Methods</h3><div>1083 adolescents (10-17 years) participated in a representative online study. The associations between PVS and ER difficulties, BW frequency, and video format preference were investigated cross-sectionally and prospectively after 10 months (n = 559) by applying logistic regression modeling. PVS classification was based on ICD-11 criteria for disorders due to addictive behaviors. The robustness of results was tested using multiple imputation.</div></div><div><h3>Results</h3><div>Frequent BW, a preference for short videos, difficulties engaging in goal-directed behavior and lack of emotional clarity were significantly associated to more adolescent PVS in the cross-sectional analysis. Strikingly, lack of emotional clarity prevailed as the only significant predictor for adolescent PVS in the longitudinal analysis, even after controlling for baseline PVS.</div></div><div><h3>Conclusion</h3><div>In line with other forms of PIU, studies on the etiology, prevention, and treatment of PVS should include frequent BW and lack of emotional clarity as crucial correlates. However, the preferred video format, especially short videos, should be considered when addressing adolescents affected by PVS. Further longitudinal studies are necessary to explore causal relationships and derive strategies that promote healthy video viewing habits.</div></div>","PeriodicalId":72681,"journal":{"name":"Computers in human behavior reports","volume":"23 ","pages":"Article 101247"},"PeriodicalIF":5.5,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148736331","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":"Technology that listens, privacy that protects: A multi-method study of voice-based digital assistants- a moderated mediation perspective","authors":"Ghazanfar Ali Abbasi","doi":"10.1016/j.chbr.2026.101242","DOIUrl":"10.1016/j.chbr.2026.101242","url":null,"abstract":"<div><div>This study investigates the factors influencing the continued use of voice-based digital assistants (VBDAs) by integrating constructs from the unified theory of acceptance and use of technology (UTAUT) and communication privacy management theory (CPMT), alongside trust. Previous research has examined technological adoption and privacy concerns in isolation. This study bridges this gap by offering a unified perspective that accounts for both functional and privacy-related drivers of VBDA usage. 487 Saudi university students with prior VBDA experience were analysed using integrated generalized structured component analysis and necessary condition analysis. The results revealed that desire for privacy significantly increases online privacy concern, whereas perceived control reduces it. Online privacy concern negatively influences continued usage intention. Conversely, trust, effort expectancy, performance expectancy, facilitating conditions, and social influence positively predict continued use. Importantly, the moderated mediation analysis showed that the indirect effect of the desire for privacy on usage intention through privacy concerns is amplified among individuals with high uncertainty avoidance. This study aligns with the growing literature connecting UTAUT and CPMT to trust as an intervening variable and extends this literature by incorporating moderated mediation and necessary condition analysis to provide a richer understanding of use behaviour towards VBDAs.</div></div>","PeriodicalId":72681,"journal":{"name":"Computers in human behavior reports","volume":"23 ","pages":"Article 101242"},"PeriodicalIF":5.5,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148736334","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":"A scoping review on behavioral, subjective, and physiological approaches towards human detection of natural vs synthetic stimuli: State of the art and quality assessment","authors":"César Villacura-Herrera,&nbsp;Frank T.J.M. Zaal,&nbsp;Ralf F.A. Cox","doi":"10.1016/j.chbr.2026.101221","DOIUrl":"10.1016/j.chbr.2026.101221","url":null,"abstract":"<div><div>Recent advances in artificial intelligence, along with the increasing availability of generative models, have enabled the production of highly realistic synthetic content across multiple modalities. The increasing realism of these synthetic outputs raises important questions regarding the human ability to detect them, representing a new challenge for human perception at multiple levels. Addressing this requires an integrative approach focused not only on behavioral performance, but also subjective experience and physiological indicators. The present review synthesizes the empirical evidence on human detection of natural and synthetic stimuli across images, text, audio and video, particularly examining responses at behavioral, subjective and physiological levels, while also providing an examination of the overall quality of current research efforts in this field. A systematic search was conducted across five databases following the PRISMA-ScR guidelines, identifying a total of 97 eligible studies. Across the literature, most studies have focused on behavioral approaches (94.85%), consistently revealing that detection performance is often near chance level in particular for highly realistic visual stimuli. A strong dissociation between accuracy rates and subjective confidence is also frequently reported, with these evaluations being shaped by familiarity and contextual cues. Physiological evidence further suggests that natural and synthetic stimuli can produce distinguishable patterns of gaze and neural activity, even when behavioral accuracy does not reflect these differences. Overall, the literature shows high methodological heterogeneity and limitations in stimulus control and reporting practices. These findings highlight the need for integrative approaches to better understand human detection and inform strategies for improving judgements in digital environments.</div></div>","PeriodicalId":72681,"journal":{"name":"Computers in human behavior reports","volume":"23 ","pages":"Article 101221"},"PeriodicalIF":5.5,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148736337","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":"Forming and losing trust with AI agents: A computational account of human–AI interaction in the trust game","authors":"Yoshiko Arima,&nbsp;Susumu Ohnuma","doi":"10.1016/j.chbr.2026.101230","DOIUrl":"10.1016/j.chbr.2026.101230","url":null,"abstract":"<div><div>How do humans form trust with AI agents? We address this question with a computational model of human trust game behaviour grounded in the Active Inference Framework (AIF), tested against two behavioural experiments with bot partners. The model introduces a parameter <span><math><mi>γ</mi></math></span>, a trial-by-trial weighting of social versus economic inference in the free energy computation; higher <span><math><mi>γ</mi></math></span> indicates an internal model in which both self and other are predicted to prioritize social over economic value. Participants’ <span><math><mi>γ</mi></math></span> was estimated from observed behaviour via gradient descent. Across both experiments, the model achieved high predictive accuracy for trust/betrayal choices (F1 <span><math><mo>&gt;</mo></math></span> .80) under cooperative and betrayal conditions alike. Study 1, in which bots maintained a fixed cooperation or betrayal pattern, replicated the simulated belief trajectories with only a small effect of Bot condition on <span><math><mi>γ</mi></math></span>. Study 2 used a between-subjects design contrasting a Trust condition (continuous cooperation) with a Betray condition (cooperation–betrayal–cooperation). The simulation predicted that <span><math><mi>γ</mi></math></span> would rebound once cooperation resumed; in human data it did not. Beliefs (posterior predictions of the partner’s next cooperation) recovered partially in the final phase, consistent with precision convergence in Bayesian updating, but investment rate and <span><math><mi>γ</mi></math></span> remained suppressed. Thus humans update their predictions about a bot’s behaviour in line with Active Inference accounts, yet their underlying social–economic value weighting carries a longer trace of the betrayal. This asymmetry, observed only in humans, points to factors beyond the present model, including individual differences such as sex. These results offer an empirical window on the human side of bidirectional human–AI alignment: once an AI partner’s behaviour has diverged from human expectations, renewed cooperative interaction alone may not be enough to restore alignment on the human side.</div></div>","PeriodicalId":72681,"journal":{"name":"Computers in human behavior reports","volume":"23 ","pages":"Article 101230"},"PeriodicalIF":5.5,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148843346","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":"Beyond screen time: Pre-sleep media use profiles and sleep outcomes in adults","authors":"Ajar Diushekeeva,&nbsp;Claudia Picard-Deland,&nbsp;Santiago Hidalgo,&nbsp;Antonio Zadra","doi":"10.1016/j.chbr.2026.101209","DOIUrl":"10.1016/j.chbr.2026.101209","url":null,"abstract":"<div><div>Pre-sleep media use has been widely implicated in sleep disturbances, yet most research has focused on children, adolescents, or young adults, and has relied on coarse measures of overall screen time while offering limited insight into patterns of use. The present cross-sectional study aimed to provide a comprehensive and nuanced assessment of media use within the hour before sleep in a large and diverse adult sample, and to examine how distinct media devices, activities, and usage patterns relate to sleep outcomes. A total of 722 adults (aged 18–75+) completed an online questionnaire assessing pre-sleep media use, media-related habits, self-reported sleep quality, sleep duration, and sleep onset latency. Hierarchical ordinal logistic regressions indicated that pre-sleep media use explained a modest but meaningful proportion of variance in sleep outcomes (approximately 8% for sleep quality and duration and 15% for sleep onset latency), with media use habits more consistently associated with poorer sleep outcomes than device type or specific media activities. Latent class analysis identified four distinct pre-sleep media use profiles: Versatile Moderate Users, Social Media-Streaming Moderate Users, Traditional Low Users, and Heavy Users. Class membership was strongly associated with sleep outcomes: Traditional Low Users exhibited consistently better sleep quality, longer sleep duration, and shorter sleep onset latency, whereas Social Media-Streaming Moderate Users and Heavy Users showed poorer sleep across outcomes. These findings underscore the importance of considering not only the amount of media use before sleep, but also the configurations of media activities and habitual patterns of engagement when evaluating associations with adult sleep health.</div></div>","PeriodicalId":72681,"journal":{"name":"Computers in human behavior reports","volume":"23 ","pages":"Article 101209"},"PeriodicalIF":5.5,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148641833","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":"Taking candy from a robot: Attitudes toward an autonomous non-humanoid service robot at real-life events","authors":"Melitta Csepregi,&nbsp;Beáta Korcsok,&nbsp;Judit Abdai,&nbsp;Ádám Miklósi,&nbsp;Bence Ferdinandy,&nbsp;Tímea Kovács,&nbsp;Sára Roszik,&nbsp;Márta Gácsi","doi":"10.1016/j.chbr.2026.101204","DOIUrl":"10.1016/j.chbr.2026.101204","url":null,"abstract":"<div><div>As the presence of social service robots in everyday settings is increasing, it is important to understand how they shape human behaviour in real-world interactions. We aimed to investigate the behavioural differences shown by visitors towards an autonomously operating non-humanoid service robot and a human server performing the same task—offering candies. Additionally, we compared the behavioural responses to the robot with and without candies and evaluated the efficiency of the robot compared to the human server. Our findings confirmed most of our predictions: visitors were more likely to orient at the robot than the human server, but they were not more likely to withdraw from the approaching robot than from the human server. Visitors showed similar frequencies of orienting, giving way, touching, pointing, and waving toward the robot, regardless of whether it was carrying candies. The robot's tray ran out of candies faster than the human server's in all nine rounds, indicating a high level of robot efficiency and differences in interaction dynamics shown towards the robot and the human server. Together, these findings demonstrate that service robots may elicit distinctive patterns of attention and engagement in public settings, underscoring their capacity to influence human behaviour without inducing avoidance. The results represent a first step towards a more nuanced understanding of human–robot interaction in service contexts and inform the responsible integration of robots into human environments.</div></div>","PeriodicalId":72681,"journal":{"name":"Computers in human behavior reports","volume":"23 ","pages":"Article 101204"},"PeriodicalIF":5.5,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148656975","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":"Investigation of individual investors' acceptance of artificial intelligence use in financial investment decisions: The case of ChatGPT","authors":"Fatih Demir,&nbsp;Duygu Fındık-Coşkunçay","doi":"10.1016/j.chbr.2026.101223","DOIUrl":"10.1016/j.chbr.2026.101223","url":null,"abstract":"<div><div>This study investigates the factors influencing financial investors' acceptance of ChatGPT, a generative artificial intelligence tool, within the context of financial decision-making. Specifically, it examines the effects of performance expectancy, effort expectancy, social influence, facilitating conditions, and personal innovativeness on individual investors’ behavioral intention to use ChatGPT. Focusing on understanding the acceptance and usage of ChatGPT among individual investors, the research uses the unified theory of acceptance and use of technology (UTAUT) as the guiding framework. Four important constructs of UTAUT (performance expectancy, effort expectancy, social influence, and facilitating conditions) and personal innovativeness, which is an important variable in the adoption of new technologies, were investigated to understand their impact on individual investors' intentions. A total of 450 individual investors who are active investors in Türkiye participated in the study. The data was evaluated using confirmatory factor analysis and structural equation modeling techniques. The maximum likelihood estimation method was used to test the model. The main findings show that facilitating conditions (0.654) impacted behavioral intention the most. Performance expectancy (0.356) and social influence (0.206) had an impact on behavioral intention. Conversely, the positive effect of effort expectancy and personal innovativeness on behavioral intention was not supported. Interestingly, effort expectancy showed a significant negative relationship with behavioral intention, while personal innovativeness was not a significant predictor, suggesting that ease of use and innovativeness may have played a different role in high-stakes financial contexts than in other ChatGPT adoption settings. The results increase our understanding of technology acceptance in the context of AI tools, and these inputs are important for formulating strategies that promote the effective incorporation of ChatGPT into financial investment decisions. It also underlines the need for wider adoption of artificial intelligence tools in financial investment decisions to improve the investment performance of individual investors. Although this study is limited to a sample, it is expected that the effects of ChatGPT-like technologies on financial investment decisions will be the source of further research due to their potential.</div></div>","PeriodicalId":72681,"journal":{"name":"Computers in human behavior reports","volume":"23 ","pages":"Article 101223"},"PeriodicalIF":5.5,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148657505","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}