Structural differences of dialogues between humans and dialogues between humans and neural networks

Context and relevance. The rapid development of generative neural networks, beginning in 2022, has created a situation where dialogue with a character previously considered fictional and inaccessible for communication becomes possible. Potentially, the development of these systems will allow humans to gain experience comparable to that of social communication. The existence of such experience raises the question of where the boundary lies between social and parasocial relationships. Objective: to determine the presence or absence of differences between a human's dialogue with a neural network and a human's dialogue with another human. Hypothesis. Dialogue between a human and a neural network is a social act and is structurally similar to dialogue between humans. Methods and materials. The study conducts a comparative analysis of dialogues between humans and humans, and between humans and the ChatGPT 3.5 neural network, from the perspective of the psycholinguistic structure of speech. Materials from an empirical study involving a sample of students and graduate students from various Moscow universities are provided. The study created a virtual environment for oral communication between a human and a neural network; the dialogues were recorded, transcribed (converted into text format without additional processing), and compared with human-to-human dialogues, which were also recorded and transcribed. The study analyzed eighty adjacency pairs—pairs of adjacent utterances, pairs of statements by different participants located in immediate proximity to each other during interaction, taken from six dialogues. Human-to-human dialogues were conducted among respondents aged 20 to 22, of whom two were female and two were male. Human-to-neural network dialogues were conducted among respondents aged 20 to 28, of whom two were male and four were female. The study employed the method of conversation analysis, focusing on the types of difficulties respondents experienced in dialogue. Additionally, the length of utterances was examined to compare speech structure. Results. The obtained results indicate significant differences in the structure of dialogue between a human and ChatGPT 3.5 compared to dialogue between two humans, in terms of the distribution of utterance lengths in words and the types of communicative difficulties in dialogue.

1. Барабанщиков, В.А., Селиванов, В.В. (2023). Редукция тревоги и депрессии через программы на гарнитуре виртуальной реальности высокой иммерсивности. Экспериментальная психология, 16(2), 36—48. https://doi.org/10.17759/exppsy.2023160203
   Barabanshchikov, V.A., Selivanov, V.V. (2023). Reducing Anxiety and Depression through Programs on a High Immersive Virtual Reality Headset. Experimental Psychology (Russia), 16(2), 36—48. (In Russ.). https://doi.org/10.17759/exppsy.2023160203
2. Барабанщиков, В.А., Селиванова, В.В. (Ред.). (2022). Влияние технологий виртуальной реальности высшего уровня на изменение психического в юношестве. М.: Универсум.
   Barabanshchikov, V.A., Selivanov, V.V. (Eds.). (2022). The impact of high-level virtual reality technologies on mental change in youth. Moscow: Universum. (In Russ.)
3. Селиванов, В.В., Майтнер, Л., Грибер, Ю.А. (2021). Особенности использования технологий виртуальной реальности при коррекции и лечении депрессии в клинической психологии. Клиническая и специальная психология, 10(3), 231—255. https://doi.org/10.17759/cpse.2021100312
   Selivanov, V.V., Meitner, L., Griber, Yu.A. (2021). Features of the Use of Virtual Reality Technologies in the Rehabilitation and Treatment of Depression in Clinical Psychology. Clinical Psychology and Special Education, 10(3), 231—255. (In Russ.). https://doi.org/10.17759/cpse.2021100312
4. Селиванов, В.В., Побокин, П.А. (2024). Особенности тревожности и саморегуляции психической деятельности в виртуальной среде. Экспериментальная психология, 17(1), 108—117. https://doi.org/10.17759/exppsy.2024170107
   Selivanov, V.V., Pobokin, P.A. (2024). Features of Anxiety and Self-Regulation of Mental Activity in a Virtual Environment. Experimental Psychology (Russia), 17(1), 108—117. (In Russ.). https://doi.org/10.17759/exppsy.2024170107
5. Улановский, А.М., Ерохина, Л.А., Ян., М.Д. (2017). Разговор при знакомстве: конверсационный анализ быстрых свиданий. Психология. Журнал высшей школы экономики, 14(1), 140—166.
   Ulanovsky, A.M., Erohina, L.A., Yan, M.D. (2017). The Talk during a Meeting: Conversation Analysis of Speed Dating. Psychology. Journal of the Higher School of Economics, 14(1), 140—166. (In Russ.)
6. Creswell, A., et al. (2018). Generative adversarial networks: An overview. IEEE signal processing magazine, 35(1), 53—65.
7. Gambino, A., Fox, J., Ratan, R.A. (2020). Building a stronger CASA: Extending the computers are social actors paradigm. Human-Machine Communication, 1, 71—85.
8. Hilliard, A., et al. (2024). Eliciting Big Five Personality Traits in Large Language Models: A Textual Analysis with Classifier-Driven Approach. arXiv preprint arXiv:2402.08341.
9. Latif, E., Zhai, X. (2023). Fine-tuning Chatgpt for automatic scoring. Computers and Education: Artificial Intelligence, arXiv.Org, abs/2310.10072.
10. McLaughlin, M.L., Cody, M.J. (1982). Awkward silences: Behavioral antecedents and consequences of the conversational lapse. Human communication research, 8(4), 299—316.
11. Nass, C., Steuer, J., Tauber, E.R. (1994). Computers are social actors. Proceedings of the SIGCHI conference on Human factors in computing systems, 72—78.
12. Sullivan, J.H., Warkentin, M., Wallace, L. (2021). So many ways for assessing outliers: What really works and does it matter? Journal of Business Research, 132, 530—543.
13. Zhang, W., et al. (2023). Fine-Tuning ChatGPT Achieves State-of-the-Art Performance for Chemical Text Mining. ChemRxiv.