Paweł Niszczota
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Paweł Niszczota has a total of 25 co-authors. Here are the top co-authors by number of shared publications:
- Michal Bialek
University of Wroclaw
12 shared publications Paul Conway
6 shared publicationsDamian Pieńkosz
4 shared publications- Rafał Muda
University of Wrocław, Poland
4 shared publications Dániel Kaszás
3 shared publicationsMichał Białek
University of Wrocław
3 shared publications
Latest Publications
- Open Access
- 22 October 2026
- Published by Elsevier in SSRN Electronic Journal
The Moral Character of Humans Behind AI Shapes Behavioral Trust in Large Language Models
When people can shape how an AI system treats others, with nothing to gain for themselves, does their moral character carry over into the AI’s behavior? And do those who interact with the AI act as if it does? We provide behavioral evidence on these ”upstream” and ”downstream” questions in a preregistered two-phase study (N = 1,268). In the Upstream Phase, participants completed morality-related personality measures and indicated how an AI system should behave when faced with economic trade-offs. Crucially, these participants did not benefit from the AI’s decision and thus had no incentive to have it behave selfishly. In the Downstream Phase, new participants played an incentivized one-shot Trust Game with a large language model (LLM) guided by one Upstream Phase participant; they were told whether that participant scored high or low in moral character, or were told nothing about it. Upstream Phase participants higher in moral character (high in Honesty-Humility and Altruism, low in dark traits) gave more other-regarding guidance: their moral character thus carried over into the LLM’s behavior. In the Downstream Phase, learning that the LLM’s guide was high in moral character increased trust, and learning that the guide was low in moral character decreased it; the decrease was larger than the increase. Overall, these findings show that people shape the AI they guide in their own image, even when unincentivized, and that users’ trust in human-guided AI depends not only on the system itself but also on moral inferences about the distal humans behind it.
- 22 October 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessThe Moral Character of Humans Behind AI Shapes Behavioral Trust in Large Language Models
When people can shape how an AI system treats others, with nothing to gain for themselves, does their moral character carry over into the AI’s behavior? And do those who interact with the AI act as if it does? We provide behavioral evidence on these ”upstream” and ”downstream” questions in a preregistered two-phase study (N = 1,268). In the Upstream Phase, participants completed morality-related personality measures and indicated how an AI system should behave when faced with economic trade-offs. Crucially, these participants did not benefit from the AI’s decision and thus had no incentive to have it behave selfishly. In the Downstream Phase, new participants played an incentivized one-shot Trust Game with a large language model (LLM) guided by one Upstream Phase participant; they were told whether that participant scored high or low in moral character, or were told nothing about it. Upstream Phase participants higher in moral character (high in Honesty-Humility and Altruism, low in dark traits) gave more other-regarding guidance: their moral character thus carried over into the LLM’s behavior. In the Downstream Phase, learning that the LLM’s guide was high in moral character increased trust, and learning that the guide was low in moral character decreased it; the decrease was larger than the increase. Overall, these findings show that people shape the AI they guide in their own image, even when unincentivized, and that users’ trust in human-guided AI depends not only on the system itself but also on moral inferences about the distal humans behind it.
- Open Access
Antisocial behavior towards large language model users: experimental evidence
The rapid spread of large language models (LLMs) has raised concerns about the social reactions they provoke. Prior research documents negative attitudes toward AI users, but it remains unclear whether such disapproval translates into costly action. We address this question in a two-phase online experiment (N = 491 Phase II participants; Phase I provided targets) where participants could spend part of their own endowment to reduce the earnings of peers who had previously completed a real-effort task with or without LLM support. On average, participants destroyed 36% of the earnings of those who relied exclusively on the model, with punishment increasing monotonically with actual LLM use. Disclosure about LLM use created a credibility gap: self-reported null use was punished more harshly than actual null use, suggesting that declarations of "no use" are treated with suspicion. Conversely, at high levels of use, actual reliance on the model was punished more strongly than self-reported reliance. Taken together, these findings provide the first behavioral evidence that the efficiency gains of LLMs come at the cost of social sanctions.
Open AccessAntisocial behavior towards large language model users: experimental evidence
The rapid spread of large language models (LLMs) has raised concerns about the social reactions they provoke. Prior research documents negative attitudes toward AI users, but it remains unclear whether such disapproval translates into costly action. We address this question in a two-phase online experiment (N = 491 Phase II participants; Phase I provided targets) where participants could spend part of their own endowment to reduce the earnings of peers who had previously completed a real-effort task with or without LLM support. On average, participants destroyed 36% of the earnings of those who relied exclusively on the model, with punishment increasing monotonically with actual LLM use. Disclosure about LLM use created a credibility gap: self-reported null use was punished more harshly than actual null use, suggesting that declarations of "no use" are treated with suspicion. Conversely, at high levels of use, actual reliance on the model was punished more strongly than self-reported reliance. Taken together, these findings provide the first behavioral evidence that the efficiency gains of LLMs come at the cost of social sanctions.
- Top Cited Papers
- 19 January 2025
- Published by Elsevier in Journal of Research in Personality
- Vol. 115
Large language models can replicate cross-cultural differences in personality
- 19 January 2025
- Published by Elsevier in Journal of Research in Personality
- Vol. 115
Top Cited PapersLarge language models can replicate cross-cultural differences in personality
- Open Access
- 21 February 2024
- Published by Center for Open Science in PsyArXiv
Using a foreign language increases risk-taking in medical decisions but not due to attenuated emotional responses
People take more risks when deciding in their foreign language. In three lab experiments (N=424), we explored two explanations of this phenomenon: reduced anticipation of regret, or increased accessibility of risk-increasing thoughts. Participants in native and foreign language considered two prenatal tests: a less sensitive but safe, and a more sensitive one – their task was to specify the level of risk of miscarriage of the more sensitive test while still choosing it. Three studies demonstrated greater willingness to take risks in a foreign language but this effect was not mediated by changes in the accessibility of thoughts or anticipated regret.
- 21 February 2024
- Published by Center for Open Science in PsyArXiv
Open AccessUsing a foreign language increases risk-taking in medical decisions but not due to attenuated emotional responses
People take more risks when deciding in their foreign language. In three lab experiments (N=424), we explored two explanations of this phenomenon: reduced anticipation of regret, or increased accessibility of risk-increasing thoughts. Participants in native and foreign language considered two prenatal tests: a less sensitive but safe, and a more sensitive one – their task was to specify the level of risk of miscarriage of the more sensitive test while still choosing it. Three studies demonstrated greater willingness to take risks in a foreign language but this effect was not mediated by changes in the accessibility of thoughts or anticipated regret.
- 28 December 2023
- Published by Elsevier in Ecological Economics
- Vol. 218
Hard to digest investments: People oppose investment in both conventional and cultured meat producers
- 28 December 2023
- Published by Elsevier in Ecological Economics
- Vol. 218
Hard to digest investments: People oppose investment in both conventional and cultured meat producers
- 29 November 2023
- Published by Poznań University of Economics and Business
- p. 133-150
Grzeszne spółki: synteza 15 lat badań teoretycznych i empirycznych
- 29 November 2023
- Published by Poznań University of Economics and Business
- p. 133-150
Grzeszne spółki: synteza 15 lat badań teoretycznych i empirycznych
- Open Access
- 24 October 2023
- Published by Center for Open Science in OSF Preprints
Using a foreign language increases risk tolerance, but not due to attenuated emotional responses or greater accessibility of risk-increasing thoughts
People have been observed to take more (calculated) risks when deciding in their foreign language compared to their native language. But the underlying mechanism of this phenomenon remains unclear. In three experiments involving a total of 544 lab participants, we explored two potential explanations: (1) the effect is attributed to reduced anticipation of regret, or (2) it is linked to increased accessibility of risk-increasing thoughts. Participants in both native and foreign language conditions were presented with two variants of a prenatal test: a less sensitive but safe option, and a more sensitive one. Their task was to specify the maximum level of risk of miscarriage they would tolerate for the more sensitive test while still choosing it. While all three studies consistently demonstrated greater risk tolerance in the foreign language condition, this effect was not mediated by changes in the accessibility of thoughts or in anticipated regret. Consequently, both proposed explanations fail to account for the observed results of the foreign language effect.
- 24 October 2023
- Published by Center for Open Science in OSF Preprints
Open AccessUsing a foreign language increases risk tolerance, but not due to attenuated emotional responses or greater accessibility of risk-increasing thoughts
People have been observed to take more (calculated) risks when deciding in their foreign language compared to their native language. But the underlying mechanism of this phenomenon remains unclear. In three experiments involving a total of 544 lab participants, we explored two potential explanations: (1) the effect is attributed to reduced anticipation of regret, or (2) it is linked to increased accessibility of risk-increasing thoughts. Participants in both native and foreign language conditions were presented with two variants of a prenatal test: a less sensitive but safe option, and a more sensitive one. Their task was to specify the maximum level of risk of miscarriage they would tolerate for the more sensitive test while still choosing it. While all three studies consistently demonstrated greater risk tolerance in the foreign language condition, this effect was not mediated by changes in the accessibility of thoughts or in anticipated regret. Consequently, both proposed explanations fail to account for the observed results of the foreign language effect.
- Top Cited Papers
- 19 August 2023
- Published by Elsevier in Finance Research Letters
- Vol. 58
GPT has become financially literate: Insights from financial literacy tests of GPT and a preliminary test of how people use it as a source of advice
- 19 August 2023
- Published by Elsevier in Finance Research Letters
- Vol. 58
Top Cited PapersGPT has become financially literate: Insights from financial literacy tests of GPT and a preliminary test of how people use it as a source of advice