Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
129 tokens/sec
GPT-4o
28 tokens/sec
Gemini 2.5 Pro Pro
42 tokens/sec
o3 Pro
4 tokens/sec
GPT-4.1 Pro
38 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Influence of External Information on Large Language Models Mirrors Social Cognitive Patterns (2305.04812v3)

Published 8 May 2023 in cs.CL

Abstract: Social cognitive theory explains how people learn and acquire knowledge through observing others. Recent years have witnessed the rapid development of LLMs, which suggests their potential significance as agents in the society. LLMs, as AI agents, can observe external information, which shapes their cognition and behaviors. However, the extent to which external information influences LLMs' cognition and behaviors remains unclear. This study investigates how external statements and opinions influence LLMs' thoughts and behaviors from a social cognitive perspective. Three experiments were conducted to explore the effects of external information on LLMs' memories, opinions, and social media behavioral decisions. Sociocognitive factors, including source authority, social identity, and social role, were analyzed to investigate their moderating effects. Results showed that external information can significantly shape LLMs' memories, opinions, and behaviors, with these changes mirroring human social cognitive patterns such as authority bias, in-group bias, emotional positivity, and emotion contagion. This underscores the challenges in developing safe and unbiased LLMs, and emphasizes the importance of understanding the susceptibility of LLMs to external influences.

Definition Search Book Streamline Icon: https://streamlinehq.com
References (85)
  1. Bandura, A. Social Foundations of Thought and Action: A Social Cognitive Theory. Prentice-Hall series in social learning theory (Prentice-Hall, 1986).
  2. Bandura, A. Social cognitive theory: An agentic perspective. Annual review of psychology 52, 1–26 (2001).
  3. Belief, Attitude, Intention, and Behavior: An Introduction to Theory and Research. Addison-Wesley series in social psychology (Addison-Wesley Publishing Company, 1975).
  4. Communication and persuasion: Central and peripheral routes to attitude change (Springer Science & Business Media, 2012).
  5. Bandura, A. Social cognitive theory of mass communication. Media psychology 3, 265–299 (2001).
  6. OpenAI. Introducing chatgpt. URL https://openai.com/blog/chatgpt. 2022, Nov 30.
  7. Anthropic. Model card and evaluations for claude models. URL https://www-files.anthropic.com/production/images/Model-Card-Claude-2.pdf. 2023, Jul 11.
  8. Touvron, H. et al. Llama: Open and efficient foundation language models. arXiv preprint arXiv:2302.13971 (2023).
  9. Can ai language models replace human participants? Trends in Cognitive Sciences (2023).
  10. Griffin, L. D. et al. Susceptibility to influence of large language models. arXiv preprint arXiv:2303.06074 (2023).
  11. Argyle, L. P. et al. Out of one, many: Using language models to simulate human samples. Political Analysis 31, 337–351 (2023).
  12. Bubeck, S. et al. Sparks of artificial general intelligence: Early experiments with gpt-4. arXiv preprint arXiv:2303.12712 (2023).
  13. Measuring and manipulating knowledge representations in language models. arXiv preprint arXiv:2304.00740 (2023).
  14. Zellers, R. et al. Defending against neural fake news. In Wallach, H. M. et al. (eds.) Advances in Neural Information Processing Systems 32: Annual Conference on Neural Information Processing Systems 2019, NeurIPS 2019, December 8-14, 2019, Vancouver, BC, Canada, 9051–9062 (2019). URL https://proceedings.neurips.cc/paper/2019/hash/3e9f0fc9b2f89e043bc6233994dfcf76-Abstract.html.
  15. Santurkar, S. et al. Whose opinions do language models reflect? arXiv preprint arXiv:2303.17548 (2023).
  16. Co-writing with opinionated language models affects users’ views. In Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems, 1–15 (2023).
  17. Exploring ai ethics of chatgpt: A diagnostic analysis. arXiv preprint arXiv:2301.12867 (2023).
  18. Ienca, M. Don’t pause giant ai for the wrong reasons. Nature Machine Intelligence 1–2 (2023).
  19. Adversarial attacks on deep-learning models in natural language processing: A survey. ACM Transactions on Intelligent Systems and Technology (TIST) 11, 1–41 (2020).
  20. Evaluating the robustness of neural language models to input perturbations. In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, 1558–1570 (Association for Computational Linguistics, Online and Punta Cana, Dominican Republic, 2021). URL https://aclanthology.org/2021.emnlp-main.117.
  21. Truthfulqa: Measuring how models mimic human falsehoods. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), 3214–3252 (2022).
  22. Data contamination: From memorization to exploitation. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), 157–165 (Association for Computational Linguistics, Dublin, Ireland, 2022). URL https://aclanthology.org/2022.acl-short.18.
  23. Why does chatgpt fall short in answering questions faithfully? arXiv preprint arXiv:2304.10513 (2023).
  24. Liang, P. et al. Holistic evaluation of language models. arXiv preprint arXiv:2211.09110 (2022).
  25. Wang, J. et al. On the robustness of chatgpt: An adversarial and out-of-distribution perspective. arXiv preprint arXiv:2302.12095 (2023).
  26. Shi, F. et al. Large language models can be easily distracted by irrelevant context. arXiv preprint arXiv:2302.00093 (2023).
  27. Are you in a masquerade? exploring the behavior and impact of large language model driven social bots in online social networks. arXiv preprint arXiv:2307.10337 (2023).
  28. Shao, C. et al. The spread of low-credibility content by social bots. Nature communications 9, 1–9 (2018).
  29. Large pre-trained language models contain human-like biases of what is right and wrong to do. Nature Machine Intelligence 4, 258–268 (2022).
  30. Ji, Z. et al. Survey of hallucination in natural language generation. ACM Computing Surveys 55, 1–38 (2023).
  31. Cognitive reflection correlates with behavior on twitter. Nature communications 12, 921 (2021).
  32. Is twitter an individual mass communication medium? In Society for Information Technology & Teacher Education International Conference, 1712–1717 (Association for the Advancement of Computing in Education (AACE), 2010).
  33. Self-efficacy: The exercise of control (1999).
  34. Pornpitakpan, C. The persuasiveness of source credibility: A critical review of five decades’ evidence. Journal of applied social psychology 34, 243–281 (2004).
  35. The sleeper effect in persuasion: a meta-analytic review. Psychological bulletin 130, 143 (2004).
  36. Wilder, D. A. Some determinants of the persuasive power of in-groups and out-groups: Organization of information and attribution of independence. Journal of Personality and Social Psychology 59, 1202 (1990).
  37. Self-efficacy in changing societies (1996).
  38. The elaboration likelihood model of persuasion (Springer, 1986).
  39. Universal dimensions of social cognition: Warmth and competence. Trends in cognitive sciences 11, 77–83 (2007).
  40. Taori, R. et al. Stanford alpaca: An instruction-following llama model. https://github.com/tatsu-lab/stanford_alpaca (2023).
  41. Chiang, W.-L. et al. Vicuna: An open-source chatbot impressing gpt-4 with 90%* chatgpt quality (2023). URL https://lmsys.org/blog/2023-03-30-vicuna/.
  42. Cross-stage neural pattern similarity in the hippocampus predicts false memory derived from post-event inaccurate information. Nature Communications 14, 2299 (2023).
  43. Brown, T. B. et al. Language models are few-shot learners. In Larochelle, H., Ranzato, M., Hadsell, R., Balcan, M. & Lin, H. (eds.) Advances in Neural Information Processing Systems 33: Annual Conference on Neural Information Processing Systems 2020, NeurIPS 2020, December 6-12, 2020, virtual (2020). URL https://proceedings.neurips.cc/paper/2020/hash/1457c0d6bfcb4967418bfb8ac142f64a-Abstract.html.
  44. Dynamics of opinion formation under majority rules on complex social networks. Scientific reports 10, 456 (2020).
  45. Influence of information attributes on information dissemination in public health emergencies. Humanities and Social Sciences Communications 9, 1–22 (2022).
  46. Emotions and information diffusion in social media—sentiment of microblogs and sharing behavior. Journal of management information systems 29, 217–248 (2013).
  47. What makes people share political content on social media? the role of emotion, authority and ideology. Computers in Human Behavior 129, 107150 (2022).
  48. Hu, E. J. et al. Lora: Low-rank adaptation of large language models. In The Tenth International Conference on Learning Representations, ICLR 2022, Virtual Event, April 25-29, 2022 (OpenReview.net, 2022). URL https://openreview.net/forum?id=nZeVKeeFYf9.
  49. OpenAI. Model index for researchers. URL https://platform.openai.com/docs/model-index-for-researchers.
  50. Ouyang, L. et al. Training language models to follow instructions with human feedback. Advances in Neural Information Processing Systems 35, 27730–27744 (2022).
  51. Turner, J. C. The experimental social psychology of intergroup behavior. Intergroup behaviour 66 (1981).
  52. Turner, J. C. et al. Social identification and psychological group formation. The social dimension: European developments in social psychology 2, 518–538 (1984).
  53. Predicting positive user responses to social media advertising: The roles of emotional appeal, informativeness, and creativity. International Journal of Information Management 36, 360–373 (2016).
  54. Twitter sentiment analysis using natural language toolkit and vader sentiment. In Proceedings of the international multiconference of engineers and computer scientists, vol. 122, 16 (2019).
  55. Clustering and sentiment analysis on twitter data. In 2017 2nd International Conference on Telecommunication and Networks (TEL-NET), 1–5 (IEEE, 2017).
  56. The psychological meaning of words: Liwc and computerized text analysis methods. Journal of language and social psychology 29, 24–54 (2010).
  57. The development and psychometric properties of liwc2015. Tech. Rep. (2015).
  58. Heuristic processing can bias systematic processing: effects of source credibility, argument ambiguity, and task importance on attitude judgment. Journal of personality and social psychology 66, 460 (1994).
  59. Blass, T. The milgram paradigm after 35 years: Some things we now know about obedience to authority 1. Journal of applied social psychology 29, 955–978 (1999).
  60. Brewer, M. B. In-group bias in the minimal intergroup situation: A cognitive-motivational analysis. Psychological bulletin 86, 307 (1979).
  61. Tajfel, H. Human Groups and Social Categories: Studies in Social Psychology (Cambridge University Press, 1981).
  62. The pollyanna hypothesis. Journal of verbal learning and verbal behavior 8, 1–8 (1969).
  63. The Pollyanna principle: Selectivity in language, memory, and thought (Schenkman Publishing Company, 1978).
  64. Experimental evidence of massive-scale emotional contagion through social networks. Proceedings of the National academy of Sciences of the United States of America 111, 8788 (2014).
  65. Cultural differences and switching of in-group sharing behavior between an american (facebook) and a chinese (renren) social networking site. Journal of Cross-Cultural Psychology 44, 106–121 (2013).
  66. Dodds, P. S. et al. Human language reveals a universal positivity bias. Proceedings of the national academy of sciences 112, 2389–2394 (2015).
  67. Emotional contagion. Current directions in psychological science 2, 96–100 (1993).
  68. Barsade, S. G. The ripple effect: Emotional contagion and its influence on group behavior. Administrative science quarterly 47, 644–675 (2002).
  69. Dwivedi, Y. K. et al. “so what if chatgpt wrote it?” multidisciplinary perspectives on opportunities, challenges and implications of generative conversational ai for research, practice and policy. International Journal of Information Management 71, 102642 (2023).
  70. Fei, N. et al. Towards artificial general intelligence via a multimodal foundation model. Nature Communications 13, 3094 (2022).
  71. Turing, A. M. Computing Machinery and Intelligence. Mind LIX, 433–460 (1950). URL https://doi.org/10.1093/mind/LIX.236.433.
  72. Using large language models to simulate multiple humans and replicate human subject studies. In International Conference on Machine Learning, 337–371 (PMLR, 2023).
  73. Milgram, S. Behavioral study of obedience. The Journal of abnormal and social psychology 67, 371 (1963).
  74. Huang, J.-t. et al. Emotionally numb or empathetic? evaluating how llms feel using emotionbench. arXiv preprint arXiv:2308.03656 (2023).
  75. Belinkov, Y. Probing classifiers: Promises, shortcomings, and advances. Computational Linguistics 48, 207–219 (2022).
  76. Evaluating the social value of online health information for third-party patients: Is uncertainty always bad? Information Processing & Management 57, 102259 (2020).
  77. Opinion amplification causes extreme polarization in social networks. Scientific Reports 12, 18131 (2022).
  78. Memory-based model editing at scale. In Chaudhuri, K. et al. (eds.) International Conference on Machine Learning, ICML 2022, 17-23 July 2022, Baltimore, Maryland, USA, vol. 162 of Proceedings of Machine Learning Research, 15817–15831 (PMLR, 2022). URL https://proceedings.mlr.press/v162/mitchell22a.html.
  79. Editing factual knowledge in language models. In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, 6491–6506 (Association for Computational Linguistics, Online and Punta Cana, Dominican Republic, 2021). URL https://aclanthology.org/2021.emnlp-main.522.
  80. Peng, B. et al. Check your facts and try again: Improving large language models with external knowledge and automated feedback. arXiv preprint arXiv:2302.12813 (2023).
  81. Scaling down to scale up: A guide to parameter-efficient fine-tuning. arXiv preprint arXiv:2303.15647 (2023).
  82. Zheng, L. et al. Judging llm-as-a-judge with mt-bench and chatbot arena. arXiv preprint arXiv:2306.05685 (2023).
  83. Wang, Y. et al. Self-instruct: Aligning language model with self generated instructions. arXiv preprint arXiv:2212.10560 (2022).
  84. Sentiments analysis of fmri using automatically generated stimuli labels under naturalistic paradigm. Scientific Reports 13, 7267 (2023).
  85. Using language in social media posts to study the network dynamics of depression longitudinally. Nature communications 13, 870 (2022).
Citations (7)
View on Semantic Scholar

Summary

We haven't generated a summary for this paper yet.

Ai Sparkles Streamline Icon: https://streamlinehq.com Summarize Now