CoChat
Sign Up
PreprintCited by 109Social Robot Interaction and HRIDigital Mental Health InterventionsAI in Service Interactions

How AI and Human Behaviors Shape Psychosocial Effects of Extended Chatbot Use: A Longitudinal Randomized Controlled Study

Open access
Fang, Cathy Mengying, Liu, Auren R., Danry, Valdemar, Lee, Eunhae, Chan, Samantha W.T., Pataranutaporn, Pat, Maes, Pattie, Phang, Jason +3 more·2026·Research Square
DOI 10.21203/rs.3.rs-8148142/v1
Open this paper in CoChat — free account

Abstract

As people increasingly seek emotional support and companionship from AI chatbots, understanding how such interactions impact mental well-being becomes critical. We conducted a four-week randomized controlled experiment (n=981, >300k messages) to investigate how interaction modes (text, neutral voice, and engaging voice) and conversation types (open-ended, non-personal, and personal) influence four psychosocial outcomes: loneliness, social interaction with real people, emotional dependence on AI, and problematic AI usage. No significant effects were detected from experimental conditions, despite conversation analyses revealing differences in AI and human behavioral patterns across the conditions. Instead, participants who voluntarily used the chatbot more, regardless of assigned condition, showed consistently worse outcomes. Individuals' characteristics, such as higher trust and social attraction towards the AI chatbot, are associated with higher emotional dependence and problematic use. These findings raise deeper questions about how artificial companions may reshape the ways people seek, sustain, and substitute human connections.

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.21203/rs.3.rs-8148142/v1

Cites

0 citation errors · 0 not checkable

Statistics

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • Voluntary use of LLM-based chatbots predicts worse psychosocial outcomes.

    “voluntary use of LLM-CBs predicted worse psychosocial outcomes and LLM-CB dependence111.”

    From A scoping review on the mental health harms of LLM-based chatbots (2026) · cites this paper as [111]

  • In a randomized controlled trial with 981 participants, voluntary use of LLM-CBs predicted worse psychosocial outcomes and LLM-CB dependence.

    “In a randomized controlled trial with 981 participants, voluntary use of LLM-CBs predicted worse psychosocial outcomes and LLM-CB dependence111.”

    From A scoping review on the mental health harms of LLM-based chatbots (2026) · cites this paper as [111]

  • Voluntary use of LLM-based chatbots predicts LLM-CB dependence.

    “voluntary use of LLM-CBs predicted worse psychosocial outcomes and LLM-CB dependence111.”

    From A scoping review on the mental health harms of LLM-based chatbots (2026) · cites this paper as [111]

  • One experimental study did not find significant effects of the manipulation, but instead found that self-motivated users of LLM-CBs showed a higher risk of problematic use.

    “one experimental study did not find significant effects of the manipulation, but instead found that self-motivated users of LLM-CBs showed a higher risk of problematic use111”

    From A scoping review on the mental health harms of LLM-based chatbots (2026) · cites this paper as [111]

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Cites

The publisher hasn’t deposited this paper’s reference list.

Research works printed but not linked: 65

  1. [1]1. H. R. Kirk, I. Gabriel, C. Summerfield, B. Vidgen, S. A. Hale, Why human-AI relationships need socioaffective alignment. arXiv preprint arXiv:2502.02528 (2025). · arXiv
  2. [3]3. T. Koulouri, R. D. Macredie, D. Olakitan, Chatbots to support young adults' mental health: an exploratory study of acceptability. ACM Transactions on Interactive Intelligent Systems (TiiS) 12 (2), 1-39 (2022).
  3. [4]4. A.Xygkou, et al. , MindTalker: Navigating the Complexities of AI-Enhanced Social Engagement for People with Early-Stage Dementia, in Proceedings of the CHI Conference on Human Factors in Computing Systems (2024), pp. 1-15.
  4. [5]5. A. Xygkou, et al. , The' Conversation' about Loss: Understanding How Chatbot Technology was Used in Supporting People in Grief., in Proceedings of the 2023 CHI conference on human factors in computing systems (2023), pp. 1-15.
  5. [6]6. L. Ring, L. Shi, K. Totzke, T. Bickmore, Social support agents for older adults: longitudinal affective computing in the home. Journal on Multimodal User Interfaces 9 (1), 79-88 (2015).
  6. [7]7. J. De Freitas, Z. O˘ guz-U˘ guralp, A. K. U˘ guralp, S. Puntoni, AI companions reduce loneliness. Journal of Consumer Research p. ucaf040 (2025).
  7. [8]8. B. Maples, M. Cerit, A. Vishwanath, R. Pea, Loneliness and suicide mitigation for students using GPT3-enabled chatbots. npj mental health research 3 (1), 4 (2024).
  8. [9]9. L. Laestadius, A. Bishop, M. Gonzalez, D. Illenˇ c´ ık, C. Campos-Castillo, Too human and not humanenough: A grounded theory analysis of mental health harms from emotional dependence on the social chatbot Replika. New Media & Society 26 (10), 5923-5941 (2024).
  9. [12]12. P. Pataranutaporn, R. Liu, E. Finn, P. Maes, Influencing human-AI interaction by priming beliefs about AI can increase perceived trustworthiness, empathy and effectiveness. Nature Machine Intelligence 5 (10), 1076-1086 (2023).
  10. [13]13. M. Sharma, et al. , Towards understanding sycophancy in language models. arXiv preprint arXiv:2310.13548 (2023). · arXiv
  11. [14]14. H. Xia, J. Chen, Y. Qiu, P. Liu, Z. Liu, The Impact of Human-Chatbot Interaction on HumanHumanInteraction: A Substitution or Complementary Effect. International Journal of HumanComputer Interaction pp. 1-13 (2024).
  12. [15]15. N. Hickin, et al. , The effectiveness of psychological interventions for loneliness: A systematic review and meta-analysis. Clinical Psychology Review 88 , 102066 (2021).
  13. [16]16. A. R. Liu, P. Pataranutaporn, P. Maes, Chatbot companionship: a mixed-methods study of companion chatbot usage patterns and their relationship to loneliness in active users. arXiv preprint arXiv:2410.21596 (2024). · arXiv
  14. [17]17. P. Pataranutaporn, Cyborg Psychology: The Art & Science of Designing Human-AI Systems that Support Human Flourishing , Ph.D. thesis, Massachusetts Institute of Technology (2024).
  15. [18]18. J. S. Park, et al. , Generative agents: Interactive simulacra of human behavior, in Proceedings of the 36th annual acm symposium on user interface software and technology (2023), pp. 1-22.
  16. [19]19. J. S. Park, et al. , Generative agent simulations of 1,000 people. arXiv preprint arXiv:2411.10109 (2024). · arXiv
  17. [20]20. K. Seaborn, N. P. Miyake, P. Pennefather, M. Otake-Matsuura, Voice in human-agent interaction: A survey. ACM Computing Surveys (CSUR) 54 (4), 1-43 (2021).
  18. [21]21. L. Reicherts, et al. , It's good to talk: A comparison of using voice versus screen-based interactions for agent-assisted tasks. ACMTransactions on Computer-Human Interaction 29 (3), 1-41 (2022).
  19. [22]22. L. Ibrahim, S. Huang, L. Ahmad, M. Anderljung, Beyond static AI evaluations: advancing human interaction evaluations for LLM harms and risks. arXiv preprint arXiv:2405.10632 (2024). · arXiv
  20. [23]23. M. Chandra, et al. , From Lived Experience to Insight: Unpacking the Psychological Risks of Using AI Conversational Agents. arXiv preprint arXiv:2412.07951 (2024). · arXiv
  21. [25]25. P. P. Liang, et al. , Hemm: Holistic evaluation of multimodal foundation models. arXiv preprint arXiv:2407.03418 (2024). · arXiv
  22. [26]26. C.-H. Wu, G. Yao, Psychometric analysis of the short-form UCLA Loneliness Scale (ULS-8) in Taiwanese undergraduate students. Pers. Individ. Dif. 44 (8), 1762-1771 (2008).
  23. [27]27. J. Lubben, et al. , Performance of an abbreviated version of the Lubben Social Network Scale among three European community-dwelling older adult populations. The Gerontologist 46 (4), 503-513 (2006).
  24. [28]28. C. M. Sirvent-Ruiz, M. d. l. V. Moral-Jim´ enez, J. Herrero, M. Miranda-Rov´ es, F. J. Rodr´ ıguez D´ ıaz, Concept of Affective Dependence and Validation of an Affective Dependence Scale. Psychology Research and Behavior Management pp. 3875-3888 (2022).
  25. [29]29. M. W. Stevens, D. Dorstyn, P. H. Delfabbro, D. L. King, Global prevalence of gaming disorder: A systematic review and meta-analysis. Aust. N. Z. J. Psychiatry 55 (6), 553-568 (2021).
  26. [30]30. Y. Zhang, D. Zhao, J. T. Hancock, R. Kraut, D. Yang, The Rise of AI Companions: How Human-Chatbot Relationships Influence Well-Being. arXiv preprint arXiv:2506.12605 (2025). · arXiv
  27. [32]32. Q. Zhu, et al. , Effects of emotional expressiveness on voice chatbot interactions, in Proceedings of the 4th conference on conversational user interfaces (2022), pp. 1-11.
  28. [33]33. L. Ibrahim, et al. , Multi-turn evaluation of anthropomorphic behaviours in large language models. arXiv preprint arXiv:2502.07077 (2025). · arXiv
  29. [35]35. A. Barak, O. Gluck-Ofri, Degree and reciprocity of self-disclosure in online forums. CyberPsychology & Behavior 10 (3), 407-417 (2007).
  30. [36]36. N. L. Collins, S. J. Read, Adult attachment, working models, and relationship quality in dating couples. Journal of personality and social psychology 58 (4), 644 (1990).
  31. [37]37. R. M. Bagby, J. D. Parker, G. J. Taylor, The twenty-item Toronto Alexithymia Scale-I. Item selection and cross-validation of the factor structure. Journal of psychosomatic research 38 (1), 23-32 (1994).
  32. [40]40. Y. Liu-Thompkins, S. Okazaki, H. Li, Artificial empathy in marketing interactions: Bridging the human-AI gap in affective and social customer experience. Journal of the Academy of Marketing Science 50 (6), 1198-1218 (2022).
  33. [42]42. J. Meng, M. Rheu, Y. Zhang, Y. Dai, W. Peng, Mediated social support for distress reduction: AI Chatbots vs. Human. Proceedings of the ACM on Human-Computer Interaction 7 (CSCW1), 1-25 (2023).
  34. [45]45. K. K. Fitzpatrick, A. Darcy, M. Vierhile, Delivering cognitive behavior therapy to young adults with symptoms of depression and anxiety using a fully automated conversational agent (Woebot): a randomized controlled trial. JMIR mental health 4 (2), e7785 (2017).
  35. [46]46. H.-P. Lee, et al. , The impact of generative AI on critical thinking: Self-reported reductions in cognitive effort and confidence effects from a survey of knowledge workers, in Proceedings of the 2025 CHI conference on human factors in computing systems (2025), pp. 1-22.
  36. [47]47. B. Sparrow, J. Liu, D. M. Wegner, Google effects on memory: Cognitive consequences of having information at our fingertips. science 333 (6043), 776-778 (2011).
  37. [48]48. H. Allcott, L. Braghieri, S. Eichmeyer, M. Gentzkow, The welfare effects of social media. American economic review 110 (3), 629-676 (2020).
  38. [49]49. E. Kross, et al. , Facebook use predicts declines in subjective well-being in young adults. PloS one 8 (8), e69841 (2013).
  39. [50]50. C. Peng, S. Zhang, F. Wen, K. Liu, How loneliness leads to the conversational AI usage intention: The roles of anthropomorphic interface, para-social interaction. Current Psychology 44 (9), 8177-8189 (2025).
  40. [51]51. A. B. Herbener, M. F. Damholdt, Are lonely youngsters turning to chatbots for companionship? The relationship between chatbot usage and social connectedness in Danish high-school students. International Journal of Human-Computer Studies 196 , 103409 (2025).
  41. [53]53. S. S. Sundar, et al. , The MAIN model: A heuristic approach to understanding technology effects on credibility (MacArthur Foundation Digital Media and Learning Initiative Cambridge, MA) (2008).
  42. [54]54. S. S. Sundar, J. Kim, Machine heuristic: When we trust computers more than humans with our personal information, in Proceedings of the 2019 CHI Conference on human factors in computing systems (2019), pp. 1-9.
  43. [55]55. M. Chandra, et al. , Longitudinal Study on Social and Emotional Use of AI Conversational Agent. arXiv preprint arXiv:2504.14112 (2025). · arXiv
  44. [56]56. M.Rubin, et al. , Comparing the value of perceived human versus AI-generated empathy. Nature Human Behaviour pp. 1-15 (2025).
  45. [57]57. J. Shen, et al. , Empathy toward artificial intelligence versus human experiences and the role of transparency in mental health and social support chatbot design: Comparative study. JMIR Mental Health 11 (1), e62679 (2024).
  46. [58]58. L. Ibrahim, F. S. Hafner, L. Rocher, Training language models to be warm and empathetic makes them less reliable and more sycophantic. arXiv preprint arXiv:2507.21919 (2025). · arXiv
  47. [59]59. A. Hurst, et al. , Gpt-4o system card. arXiv preprint arXiv:2410.21276 (2024). · arXiv
  48. [60]60. R. D. Hays, M. R. DiMatteo, A short-form measure of loneliness. J. Pers. Assess. 51 (1), 69-81 (1987).
  49. [61]61. S.-C. Yu, H.-R. Chen, Y.-W. Yang, Development and validation the Problematic ChatGPT Use Scale: a preliminary report. Current Psychology 43 (31), 26080-26092 (2024).
  50. [63]63. D. Tingley, T. Yamamoto, K. Hirose, L. Keele, K. Imai, Mediation:R Package for causal mediation analysis. J. Stat. Softw. 59 (5), 1-38 (2014).
  51. [64]64. Y. Benjamini, Y. Hochberg, Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing. Journal of the Royal Statistical Society. Series B (Methodological) 57 (1), 289-300 (1995).
  52. [65]65. R. M. O'brien, A caution regarding rules of thumb for variance inflation factors. Qual. Quant. 41 (5), 673-690 (2007).
  53. [66]66. J. Hair, W. Black, B. Babin, R. Anderson, Multivariate Data Analysis: A Global Perspective (Pearson) (2010).
  54. [68]68. C. M. Masi, H.-Y. Chen, L. C. Hawkley, J. T. Cacioppo, A meta-analysis of interventions to reduce loneliness. Pers. Soc. Psychol. Rev. 15 (3), 219-266 (2011).
  55. [69]69. A. M. Eccles, P. Qualter, Review: Alleviating loneliness in young people - a meta-analysis of interventions. Child Adolesc. Ment. Health 26 (1), 17-33 (2021).
  56. [70]70. Q. Chang, F. Sha, C. H. Chan, P. S. F. Yip, Validation of an abbreviated version of the Lubben Social Network Scale ('LSNS-6') and its associations with suicidality among older adults in China. PLoS One 13 (8), e0201612 (2018).
  57. [71]71. T. D. Buckley, T. D. Becker, D. Burnette, Validation of the abbreviated Lubben Social Network Scale (LSNS-6) and its association with self-rated health amongst older adults in Puerto Rico. Health Soc. Care Community 30 (6), e5527-e5538 (2022).
  58. [72]72. D. Johnson, K. Grayson, Cognitive and affective trust in service relationships. Journal of Business research 58 (4), 500-507 (2005).
  59. [73]73. S. W. T. Chan, T. S. Gunasekaran, Y. S. Pai, H. Zhang, S. Nanayakkara, KinVoices: Using Voices of Friends and Family in Voice Interfaces. Proc. ACM Hum.-Comput. Interact. 5 (CSCW2), 1-25 (2021).
  60. [75]75. A. Abdulrahman, D. Richards, A. Aysin Bilgin, A comparison of human and machine-generated voice, in 25th ACM Symposium on Virtual Reality Software and Technology (ACM, New York, NY, USA) (2019).
  61. [76]76. N. I. Fisher, R. E. Kordupleski, Good and bad market research: A critical review of Net Promoter Score. Appl. Stoch. Models Bus. Ind. 35 (1), 138-151 (2019).
  62. [77]77. N. Raj Prabhu, C. Raman, H. Hung, Defining and quantifying conversation quality in spontaneous interactions, in Companion Publication of the 2020 International Conference on Multimodal Interaction (ACM, New York, NY, USA) (2020).
  63. [79]79. B. Rammstedt, O. P. John, Measuring personality in one minute or less: A 10-item short version of the Big Five Inventory in English and German. Journal of research in Personality 41 (1), 203-212 (2007).
  64. [80]80. Z. Ma, et al. , emotion2vec: Self-supervised pre-training for speech emotion representation. arXiv preprint arXiv:2312.15185 (2023). · arXiv
  65. [81]81. C. Hutto, E. Gilbert, Vader: A parsimonious rule-based model for sentiment analysis of social media text, in Proceedings of the international AAAI conference on web and social media , vol. 8 (2014), pp. 216-225.

Cited by

Sorted by citations

Related

Authors
Fang, C. M., Liu, A. R., Danry, V., Lee, E., Chan, S. W. T., Pataranutaporn, P., Maes, P., Phang, J., Lampe, M., Ahmad, L., & Agarwal, S.
Affiliations
  • MIT Media Lab (US)
  • Massachusetts Institute of Technology (US)
  • OpenAI (United States) (US)
ORCID
Venue
Research Square
Publisher
Springer Science and Business Media LLC
Published
2026-04-27
DOI
10.21203/rs.3.rs-8148142/v1
OpenAlex
W4414820664
Cited by
109
References
0
Citations by year
2026: 105 · 2025: 4
Type
posted content
Language
en
Open access
Licence
Locations
  • Research Square (acceptedVersion, cc-by)
  • arXiv (Cornell University) (submittedVersion)
  • arXiv (Cornell University) (cc-by)
Topics
  • Social Sciences › Psychology › Social Psychology › Social Robot Interaction and HRI
  • Social Sciences › Psychology › Applied Psychology › Digital Mental Health Interventions
  • Physical Sciences › Computer Science › Artificial Intelligence › AI in Service Interactions
Keywords
AI chatbots, human-AI interaction, loneliness, conversation types, trust in AI, social attraction, psychological well-being, emotional support

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review