Large language models capable of distinguishing between single and repeated gambles: Understanding and intervening in risky choice

doi:10.3724/SP.J.1041.2026.0416

Abstract

Abstract:

Risky choice (RC) is a common and important form of decision making in daily life. Its theoretical development primarily follows two major theories: normative theory and descriptive theory. The paradigms of single- and repeated-play gambles can provide an effective framework for distinguishing between the theories. However, prior research lacks direct observations of the decision-making process, which can limit the deep understanding of individual behaviour and hinder the development of effective behavioural interventions. In recent years, large language models (LLMs) have demonstrated highly human-like characteristics by not only simulating human preferences in behavioural performance but also exhibiting similar reasoning pathways. This offers a promising solution to the aforementioned limitations. This study, which is grounded in the classic RC paradigms of single versus repeated gambles, investigates the capability of LLMs to simulate and understand risk preferences and decision-making processes. Specifically, this study explores the potential of LLMs’ understanding of decision strategies to generate intervention texts and evaluates their effectiveness in influencing human decisions.

This work comprises three studies. In Study 1, GPT-3.5 and GPT-4 were employed to simulate human responses to gambling decisions under nine probability conditions (with constant expected value), which generated a total of 3, 600 responses across single and repeated gamble scenarios. In Study 2, LLM-generated strategies were constructed through a three-stage process (decision rationale extraction, strategy generation and quality evaluation), then the human participants were required to complete decision-making tasks in two experiments: Experiment 1 replicated the medical/financial scenarios (N = 349, N_male = 174, M_age = 21.79) of Sun et al. (2014) in a 2 (context: medical vs. financial) × 2 (application frequency: single vs. repeated) within-subjects design, and Experiment 2 examined digital contexts with a 2 (context: content creation vs. e-commerce marketing) × 2 (frequency: single vs. repeated) mixed design (context as between subjects). Subsequently, DeepSeek-R1 was used to perform the same tasks and generate strategy texts through the three-stage process. Finally, the participants were instructed to evaluate their acceptance of the LLM-generated strategies. Study 3 extended the Study 2 methodology to determine whether the LLM-generated intervention texts could reverse the participants’ classic choice preference across the single versus repeated gamble scenarios. The Study 2 experimental contexts (Experiment 1: medical vs. financial, N = 460, N_male = 205, M_age = 21.80; Experiment 2: content creation vs. e-commerce marketing, N = 240, N_male = 106, M_age = 29.12) were mirrored in Study 3, in which strategically designed intervention texts were presented during the decision-making tasks to test their capacity to modify the participants’ inherent risk preference between the single and repeated gamble conditions and evaluate the persuasive efficacy of LLM-generated strategies on human decision biases.

Study 1 shows that the LLMs (GPT-3.5 and GPT-4) can successfully replicate the typical human pattern of risk aversion in single-play scenarios and risk seeking in repeated-play scenarios, though both models demonstrated an overall stronger tendency toward risk seeking compared with the human participants. Study 2 demonstrates that the human participants preferred low-EV certain options in single-play contexts and high-EV risky options in repeated-play contexts in both experiments. The participants also showed high agreement with the strategies generated by the LLMs in different scenarios. Study 3 confirms that the LLM-generated intervention texts can significantly influence the participants’ choice tendency in all four scenarios, with strong intervention effects observed in the single-play contexts. The LLM intervention strategies are characterised by reliance on expected value computations (normative) when promoting RCs and emphasis on certainty and robustness (descriptive) when promoting safe choices.

In summary, this study demonstrates that (1) LLMs can effectively simulate context-dependent human preferences in RC, particularly the shift from risk aversion in single plays to risk seeking in repeated plays; (2) LLMs can distinguish between the logic underlying single and repeated gambles and apply normative and descriptive reasoning accordingly to externalise decision strategies; and (3) the decision strategies extracted from LLM-generated reasoning can be used to construct effective intervention texts that can alter human preferences in classic risk decision tasks, thereby validating the feasibility and effectiveness of an LLM-based cognitive intervention pathway. This study offers a new technological paradigm for AI-assisted decision intervention and expands the application boundary of LLMs to human cognitive process modelling and regulation.

Key words: risk decision-making, single- vs. repeated-play gambles, large language models, decision strategy, intervention

ZHOU Lei, LI Litong, WANG Xu, OU Huafeng, HU Qianyu, LI Aimei, GU Chenyan. (2026). Large language models capable of distinguishing between single and repeated gambles: Understanding and intervening in risky choice. Acta Psychologica Sinica, 58(3), 416-436.

Figures/Tables 30

Table 1 Experimental Task Parameters in Study 1

Gambling Task
Gain Outcome		Loss Outcome
Amount (CNY)	Probability (%)	Amount (CNY)	Probability (%)
+10000	10	?278	10
+5000	20	?313	20
+3333	30	?357	30
+2500	40	?417	40
+2000	50	?500	50
+1667	60	?625	60
+1429	70	?833	70
+1250	80	?1250	80
+1111	90	?2500	90

Figure 1 Example Prompts Used in Study 1

Table 2 Regression Analysis Results for Study 1

Predictor	β	SE	95% CI	Wald χ2	Exp (β)	p
Intercept	1.867	0.098	[1.679, 2.064]	362.10	6.47	< 0.001
Gamble frequency (Repeated-play = 1, Single-play = 0)	0.597	0.159	[0.289, 0.911]	14.17	1.82	< 0.001
Model type (GPT-4 = 1, GPT-3.5 = 0)	?0.815	0.124	[?1.061, ?0.574]	43.09	0.44	< 0.001
Gamble frequency × Model type	0.251	0.202	[?0.149, 0.646]	1.55	1.29	0.213

Figure 2 Proportions of Choosing to Participate in Gambling in Single-Play and Repeated-Play Gambles: LLMs vs. Humans in Study 1.
Note. A. The average proportion with which GPT (3.5/4) chose to participate in the gambling task across all probability conditions. B. The proportion with which GPT (3.5/4) chose to participate when the winning probability was 50%, together with the corresponding human data reported under the same probability condition in Redelmeier and Tversky (1992). Within each set of bars, from left to right, the bars represent GPT-3.5, GPT-4, and the human group, respectively.

Figure 3 Workflow for LLM-Generated Decision Strategies.

Figure 4 Example Prompt for Inductive Thematic Analysis in Study 2.

Figure 5 Examples of the Experimental Tasks in Study 2 and LLM-Generated Decision Strategies

Table 3 Cumulative Link Mixed Model (CLMM) Regression Results for the Effects of Gamble Frequency, Decision Context, and Their Interaction on Risky Choice

Predictor	β	SE	z	95% CI	p
Gamble frequency (Repeated-play = 1, Single-play = 0)	1.967	0.175	11.245	[1.624, 2.309]	< 0.001
Decision context (Financial = 1, Medical = 0)	?1.828	0.242	?7.549	[?2.303, ?1.354]	< 0.001
Gamble frequency × Decision context	0.766	0.265	2.886	[0.246, 1.286]	0.004

Figure 6 Choice distributions in single-play and repeated-play gamble across the medical and financial decision contexts in Study 2, Experiment 1.
Note. The vertical axis represents responses on a 4-point rating scale, where Option A denotes the certain option and Option B denotes the risky option. Specifically, “1” indicates very likely to choose Option A, “2” indicates likely to choose Option A, “3” indicates likely to choose Option B, and “4” indicates very likely to choose Option B. In each boxplot, horizontal lines from top to bottom represent the upper quartile, median, and lower quartile, respectively; overlapping lines indicate identical statistical values. White dots denote the mean (M), and error bars represent the standard error (SE). Within each pair of distributions, the left and right panels correspond to the single-play gamble and repeated-play gamble conditions, respectively.

Figure 7 Distribution of participants’ ratings of decision-strategy texts across conditions in Study 2, Experiment 1.
Note. The horizontal axis represents the perceived similarity between the decision-strategy text and participants’ own reasoning paths during the decision process. Ratings were collected on a 7-point Likert scale (1 = completely dissimilar, 7 = completely similar). For each score, bars from left to right correspond to the medical-single-play, medical-repeated-play, financial-single-play, financial-repeated-play, and overall conditions (the aggregated distribution across all four contexts).

Table 4 Cumulative Link Mixed Model (CLMM) Regression Results for the Effects of Gamble Frequency, Decision Context, and Their Interaction on Risky Choice

Predictor	β	SE	z	95% CI	p
Gamble frequency (Repeated-play = 1, Single-play = 0)	0.635	0.263	2.415	[0.120, 1.151]	0.016
Decision context (Content = 1, E-commerce = 0)	0.619	0.273	2.272	[0.085, 1.153]	0.023
Gamble frequency × Decision context	?0.003	0.362	?0.009	[?0.713, 0.706]	0.993

Figure 8 Choice distributions in single-play and repeated-play gambles across the content creation and e-commerce marketing contexts in Study 2, Experiment 2
Note. The vertical axis represents responses on a 4-point rating scale, where Option A denotes the certain option and Option B denotes the risky option. Specifically, “1” indicates very likely to choose Option A, “2” indicates likely to choose Option A, “3” indicates likely to choose Option B, and “4” indicates very likely to choose Option B. In each boxplot, horizontal lines from top to bottom represent the upper quartile, median, and lower quartile, respectively; overlapping lines indicate identical statistical values. White dots denote the mean (M), and error bars represent the standard error (SE). Within each pair of distributions, the left and right panels correspond to the single-play gamble and repeated-play gamble conditions, respectively.

Figure 9 Distribution of participants’ ratings of decision-strategy texts across conditions in Study 2, Experiment 2
Note. The horizontal axis represents the perceived similarity between the decision-strategy text and participants’ own reasoning paths during the decision process. Ratings were collected on a 7-point Likert scale (1 = completely dissimilar, 7 = completely similar). For each score, bars from left to right correspond to the content creation-single-play, content creation- repeated-play, e-commerce marketing-single-play, e-commerce marketing- repeated-play, overall content creation, and overall e-commerce marketing, respectively. The overall conditions represent pooled distributions combining the two gamble types within each context.

Table 5 Cumulative Link Mixed Model (CLMM) Regression Results for the Effects of Text Type, Gamble Frequency, Decision Context, and Their Interactions on Risky Choice

Predictor	β	SE		z	95% CI	p
Text type (Intervention = 1, Control = 0)	1.072	0.180		5.940	[0.718, 1.426]	< 0.001
Gamble frequency (Repeated-play = 1, Single-play = 0)	1.500	0.203		7.397	[1.103, 1.898]	< 0.001
Decision context (Financial = 1, Medical = 0)	?1.518	0.232		?6.530	[?1.974, ?1.062]	< 0.001
Text type × Gamble frequency	?1.840	0.284		?6.480	[?2.397, ?1.284]	< 0.001
Text type × Decision context	0.096	0.283		0.339	[?0.458, 0.650]	0.735
Gamble frequency × Decision context	0.503	0.288		1.749	[?0.061, 1.067]	0.080
Text type × Gamble frequency × Decision context	?0.220	0.384	?0.573		[?0.972, 0.532]	0.567

Figure 10 Choice distributions in single-play and repeated-play gamble tasks across medical and financial decision contexts in Study 3, Experiment 1
Note. In Panels A and B, the vertical axis represents choice ratings measured on a 4-point scale, where Option A denotes the certain option and Option B denotes the risky option. Specifically, 1 indicates very likely to choose Option A, 2 indicates likely to choose Option A, 3 indicates likely to choose Option B, and 4 indicates very likely to choose Option B. In the boxplots, the horizontal lines from top to bottom represent the upper quartile, median, and lower quartile, respectively; overlapping lines indicate identical statistics. White dots denote the mean (M), and error bars represent the standard error (SE). Within each pair of distributions, bars from left to right correspond to the control group and the intervention group, respectively.

Table 6 Cumulative Link Mixed Model (CLMM) Regression Results for the Effects of Text Type, Gamble Frequency, Decision Context, and Their Interactions on Risky Choice

Predictor	β	SE	z	95% CI	p
Text type (Intervention = 1, Control = 0)	0.961	0.335	2.867	[0.304, 1.619]	0.004
Gamble frequency (Repeated-play = 1, Single-play = 0)	0.819	0.350	2.341	[0.133, 1.505]	0.019
Decision context (Content = 1, E-commerce = 0)	0.917	0.335	2.735	[0.26, 1.574]	0.006
Text type × Gamble frequency	?2.101	0.487	?4.314	[?3.056, ?1.147]	< 0.001
Text type × Decision context	?0.266	0.464	?0.572	[?1.176, 0.645]	0.567
Gamble frequency × Decision context	?0.534	0.471	?1.132	[?1.457, 0.39]	0.257
Text type × Gamble frequency × Decision context	0.070	0.666	0.105	[?1.235, 1.375]	0.916

Figure 11 Choice distributions in single-play and repeated-play gamble tasks across content creation and e-commerce marketing contexts in Study 3, Experiment 2.
Note. In Panels A and B, the vertical axis represents choice ratings measured on a 4-point scale, where Option A denotes the certain option and Option B denotes the risky option. Specifically, 1 indicates very likely to choose Option A, 2 indicates likely to choose Option A, 3 indicates likely to choose Option B, and 4 indicates very likely to choose Option B. In the boxplots, the horizontal lines from top to bottom represent the upper quartile, median, and lower quartile, respectively; overlapping lines indicate identical statistics. White dots denote the mean (M), and error bars represent the standard error (SE). Within each pair of distributions, bars from left to right correspond to the control group and the intervention group, respectively.

Figure S1 Questionnaire flowchart for Study 2, Experiment 1.

Figure S2 Questionnaire flowchart for Study 2, Experiment 2.

Figure S3 Questionnaire flowchart for Study 3, Experiment 1.

Figure S4 Questionnaire flowchart for Study 3, Experiment 2.

Table S1 Comparison Between DeepSeek-R1 and Other Leading Large Language Models

Benchmark (Metric)		Claude-3.5- Sonnet-1022	GPT-4o 0513	Deep Seek-V3	OpenAI o1-mini	OpenAI o1-1217	DeepSeek- R1
Mathematics	AIME2024_(Pass@1)	16.0	9.3	39.2	63.6	79.2	79.8
Mathematics	MATH-500_(Pass@1)	78.3	74.6	90.2	90.0	96.4	97.3
Programming	LiveCodeBench (_Pass@1-COT)	38.9	32.9	36.2	53.8	63.4	65.9
Programming	Codeforces (_Percentile)	20.3	23.6	58.7	93.4	96.6	96.3
Open-ended Tasks	AlpacaEval2.0_(LC-winrate)	52.0	51.1	70.0	57.8	-	87.6
Open-ended Tasks	ArenaHard (_GPT-4-1106)	85.2	80.4	85.5	92.0	-	92.3
Chinese Benchmarks	CLUEWSC (_EM)	85.4	87.9	90.9	89.9	-	92.8
Chinese Benchmarks	C-Eval (_EM)	76.7	76.0	86.5	68.9	-	91.8

Table S2 Demographic Distribution of Participants in the Content Evaluation

Variable	Category	Percentage (%)	Variable	Category	Percentage (%)
Age	≤ 19	6.67/7.32	Education level	Associate degree or below	6.67/14.63
	20 ~ 22	26.67/17.07		Bachelor’ s degree	80.00/73.17
	23 ~ 25	9.99/4.88		Master’ s degree or above	13.33/12.20
	≥ 26	56.67/70.73	Field of study	Law	10.00/2.44
Gender	Male	36.67/43.90		Economics & Management	36.67/31.70
Gender	Female	63.33/56.10		Science & Engineering	43.33/60.98
Years of work experience	≤ 2 years	20.00/17.65		Medicine	3.33/2.44
	3 ~ 5 years	15.00/20.59		Arts	6.67/2.44
	≥ 6 years	65.00/61.76		Other	0.00/0.00

Table S3 Results of the Content Evaluation

Variable	t	p	Cohen’ d	M	SD
Overall rating	39.48/68.91	< 0.001/< 0.001	7.21/10.76	5.38/5.49	0.26/0.18
Reasonableness	19.81/32.72	< 0.001/< 0.001	3.62/5.20	5.40/5.20	0.52/0.33
Professionalism	20.14/32.44	< 0.001/< 0.001	3.68/5.31	5.46/5.31	0.53/0.36
Logical coherence	16.52/43.91	< 0.001/< 0.001	3.02/5.48	5.34/5.48	0.61/0.29
Readability	16.39/42.64	< 0.001/< 0.001	2.99/5.64	5.42/5.64	0.64/0.32
Persuasiveness	17.64/43.27	< 0.001/< 0.001	3.22/5.81	5.30/5.81	0.56/0.34

Table S4 Demographic Characteristics of Participants in Study 2, Experiment 1

Variable	Category	Percentage (%)	Variable	Category	Percentage (%)
Age	≤ 19	9.74	Education level	Senior high school / technical secondary school / vocational school	1.15
	20 ~ 22	57.02		Junior college	6.59
	23 ~ 25	32.09		Bachelor’ s degree	75.36
	≥ 26	1.15		Master’ s degree or above	16.90
Gender	Male	49.86	Field of study	Law	6.31
Gender	Female	50.14		Economics & Management	32.66
Monthly income (RMB)	≤ 1,000	5.73		Science & Engineering	47.85
	1,001 ~ 1,500	35.24		Medicine	6.30
	1,501 ~ 2,000	28.08		Arts	5.73
	≥ 2,001	30.95		Other	1.15

Table S5 Participants’ Ratings of Decision-Strategy Texts in Study 2, Experiment 1

Condition	t(df)	p	Cohen’ d	M	SD
Overall	t(348) = 63.04	< 0.001	3.37	5.97	0.73
Medical context - Single-play gamble	t(348) = 37.42	< 0.001	2.00	5.89	1.19
Medical context - Repeated-play gamble	t(348) = 36.00	< 0.001	1.93	5.77	1.18
Financial context - Single-play gamble	t(348) = 45.87	< 0.001	2.46	6.17	1.09
Financial context - Repeated-play gamble	t(348) = 46.16	< 0.001	2.47	6.05	1.03

Table S6 Demographic Characteristics of Participants in Study 2, Experiment 2

Variable	Category	Percentage (%)	Variable	Category	Percentage (%)
Age	≤ 19	6.19	Education level	Associate degree or below	12.86
	20 ~ 22	25.71		Bachelor’ s degree	69.52
	23 ~ 25	18.57		Master’ s degree or above	17.62
	≥ 26	49.52	Field of study	Law	5.71
Gender	Male	43.81		Economics & Management	38.57
Gender	Female	56.19		Science & Engineering	44.29
Years of work experience	≤ 2 years	36.67		Medicine	4.29
	3 ~ 5 years	20.00		Arts	5.71
	≥ 6 years	43.33		Other	1.43

Table S7 Participants’ Ratings of Decision-Strategy Texts in Study 2, Experiment 2

Condition	t(df)	p	Cohen’ d	M	SD
Overall	t(209) = 41.78	< 0.001	2.88	5.99	0.86
Content creation - Single-play gamble	t(104) = 23.12	< 0.001	2.26	5.88	1.05
Content creation - Repeated-play gamble	t(104) = 19.64	< 0.001	2.00	5.75	1.17
E-commerce marketing - Single-play gamble	t(104) = 25.14	< 0.001	2.45	6.15	1.08
E-commerce marketing - Repeated-play gamble	t(104) = 29.60	< 0.001	2.89	6.18	0.93

Table S8 Demographic Characteristics of Participants in Study 3, Experiment 1

Variable	Category	Percentage (%)	Variable	Category	Percentage (%)
Age	≤ 19	10.22	Education level	Senior high school / technical secondary school / vocational school	2.17
	20 ~ 22	54.78		Junior college	6.09
	23 ~ 25	34.78		Bachelor’ s degree	72.61
	≥ 26	0.22		Master’ s degree or above	19.13
Gender	Male	44.57	Field of study	Law	7.61
Gender	Female	55.47		Economics & Management	30.22
Monthly income (RMB)	≤ 1,000	11.52		Science & Engineering	45.65
	1,001 ~ 1,500	28.91		Medicine	7.61
	1,501 ~ 2,000	29.57		Arts	6.74
	≥ 2,001	30.00		Other	2.17

Table S9 Demographic Characteristics of Participants in Study 3, Experiment 2

Variable	Category	Percentage (%)	Variable	Category	Percentage (%)
Age	≤ 19	5.42	Education level	Associate degree or below	10.42
	20 ~ 22	24.58		Bachelor’ s degree	75.83
	23 ~ 25	16.67		Master’ s degree or above	13.75
	≥ 26	53.33	Field of study	Law	6.25
Gender	Male	44.17		Economics & Management	38.75
Gender	Female	55.83		Science & Engineering	45.42
Years of work experience	≤ 2 years	16.03		Medicine	4.17
	3 ~ 5 years	20.51		Arts	4.58
	≥ 6 years	63.46		Other	0.83

References 93

[1]	Achiam J., Adler S., Agarwal S., Ahmad L., Akkaya I., Aleman F. L.,... McGrew B.(2023). GPT-4 technical report. arXiv preprint. https://doi.org/10.48550/arXiv.2303.08774
[2]	Aher G. V., Arriaga R. I., & Kalai A. T. (2023). Using large language models to simulate multiple humans and replicate human subject studies. In Proceedings of the 40th International Conference on Machine Learning (pp. 337-371). PMLR. https://proceedings.mlr.press/v202/aher23a.html
[3]	Altay S., Hacquin A. S., Chevallier C., & Mercier H. (2023). Information delivered by a chatbot has a positive impact on COVID-19 vaccines attitudes and intentions. Journal of Experimental Psychology: Applied, 29(1), 52-62. https://doi.org/10.1037/xap0000400 doi: 10.1037/xap0000400 URL
[4]	Anderson M. A. B., Cox D. J., & Dallery J. (2023). Effects of economic context and reward amount on delay and probability discounting. Journal of the Experimental Analysis of Behavior, 120(2), 204-213. https://doi.org/10.1002/jeab.86 doi: 10.1002/jeab.868 URL pmid: 37311053
[5]	Argyle L. P., Busby E. C., Fulda N., Gubler J., Rytting C., & Wingate D. (2023). Out of one, many: Using language models to simulate human samples. Political Analysis, 31(3), 337-351. https://doi.org/10.1017/pan.2023.2 doi: 10.1017/pan.2023.2 URL
[6]	Arora C., Sayeed A. I., Licorish S., Wang F., & Treude C. (2024). Optimizing large language model hyperparameters for code generation. arXiv preprint. https://doi.org/10.48550/arXiv.2408.10577
[7]	Barberis N., & Huang M. (2009). Preferences with frames: A new utility specification that allows for the framing of risks. Journal of Economic Dynamics and Control, 33(8), 1555-1576. https://doi.org/10.1016/j.jedc.2009.01.009 doi: 10.1016/j.jedc.2009.01.009 URL
[8]	Benartzi S., & Thaler R. H. (1999). Risk aversion or myopia? Choices in repeated gambles and retirement investments. Management Science, 45(3), 364-381. https://doi.org/10.1287/mnsc.45.3.364 doi: 10.1287/mnsc.45.3.364 URL
[9]	Binz M., & Schulz E. (2023). Using cognitive psychology to understand GPT-3. Proceedings of the National Academy of Sciences, 120(6), e2218523120. https://doi.org/10.1073/pnas.2218523120
[10]	Brandstätter E., Gigerenzer G., & Hertwig R. (2006). The priority heuristic: Making choices without trade-offs. Psychological Review, 113(2), 409-432. https://doi.org/10.1037/0033-295X.113.2.409 doi: 10.1037/0033-295X.113.2.409 URL pmid: 16637767
[11]	Brislin, R. W. (1986). The wording and translation of research instruments. In W. J. Lonner & (Eds.), Field methods in cross-cultural research (pp. 137-164). Sage Publications.
[12]	Brown T., Mann B., Ryder N., Subbiah M., Kaplan J. D., Dhariwal P., … Amodei D. (2020). Language models are few-shot learners. Advances in Neural Information Processing Systems, 33, 1877-1901. https://doi.org/10.48550/arXiv.2005.14165
[13]	Carvalho T., Negm H., & El-Geneidy A. (2024). A comparison of the results from artificial intelligence-based and human-based transport-related thematic analysis. Findings. https://doi.org/10.32866/001c.94401
[14]	Chen Y., Liu T. X., Shan Y., & Zhong S. (2023). The emergence of economic rationality of GPT. Proceedings of the National Academy of Sciences, 120(51), e2316205120. https://doi.org/10.1073/pnas.2316205120
[15]	Choi S., Kang H., Kim N., & Kim J. (2025). How does artificial intelligence improve human decision-making? Evidence from the AI-powered Go program. Strategic Management Journal, 46(6), 1523-1554. https://doi.org/10.1002/smj.3694 doi: 10.1002/smj.v46.6 URL
[16]	Christensen R. H. B. (2023). ordinal: Regression models for ordinal data (R package version 2023.12-4.1) [Computer software]. https://CRAN.R-project.org/package=ordinal
[17]	Coda-Forno J., Witte K., Jagadish A. K., Binz M., Akata Z., & Schulz E. (2023). Inducing anxiety in large language models can induce bias. arXiv preprint. https://doi.org/10.48550/arXiv.2304.11111
[18]	Dai S. C., Xiong A., & Ku L. W. (2023). LLM-in-the-loop: Leveraging large language model for thematic analysis. arXiv preprint. https://doi.org/10.48550/arXiv.2310.15100
[19]	de Kok T. (2025). ChatGPT for textual analysis? How to use generative LLMs in accounting research. Management Science, 71(9), 7888-7906. https://doi.org/10.1287/mnsc.2023.03253 doi: 10.1287/mnsc.2023.03253 URL
[20]	de Varda A. G., Saponaro C., & Marelli M. (2025). High variability in LLMs’ analogical reasoning. Nature Human Behaviour, 9(7), 1339-1341. https://doi.org/10.1038/s41562-025-02224-3 doi: 10.1038/s41562-025-02224-3 URL
[21]	DeepSeek-AI Guo, D. Yang, D. Zhang, H. Song, J. Zhang, R., … Zhang Z. (2025). Deepseek-R1: Incentivizing reasoning capability in LLMs via reinforcement learning. arXiv preprint. https://doi.org/10.48550/arXiv.2501.12948
[22]	Deiana G., Dettori M., Arghittu A., Azara A., Gabutti G., & Castiglia P. (2023). Artificial intelligence and public health: Evaluating ChatGPT responses to vaccination myths and misconceptions. Vaccines, 11(7), 1217. https://doi.org/10.3390/vaccines11071217 doi: 10.3390/vaccines11071217 URL
[23]	Deiner M. S., Honcharov V., Li J., Mackey T. K., Porco T. C., & Sarkar U. (2024). Large language models can enable inductive thematic analysis of a social media corpus in a single prompt: Human validation study. JMIR Infodemiology, 4(1), e59641. https://doi.org/10.2196/59641 doi: 10.2196/59641 URL
[24]	Demszky D., Yang D., Yeager D. S., Bryan C. J., Clapper M., Chandhok S.,... Pennebaker J. W. (2023). Using large language models in psychology. Nature Reviews Psychology, 2(11), 688-701. https://doi.org/10.1038/s44159-023-00241-5
[25]	Dillion D., Tandon N., Gu Y., & Gray K. (2023). Can AI language models replace human participants? Trends in Cognitive Sciences, 27(7), 597-600. https://doi.org/10.1016/j.tics.2023.04.008 doi: 10.1016/j.tics.2023.04.008 URL pmid: 37173156
[26]	Ding Y., Zhang L. L., Zhang C., Xu Y., Shang N., Xu J., Yang F., & Yang M. (2024). Longrope: Extending LLM context window beyond 2 million tokens. arXiv preprint. https://doi.org/10.48550/arXiv.2402.13753
[27]	Faul F., Erdfelder E., Lang A. G., & Buchner A. (2007). GPower 3: A flexible statistical power analysis program for the social, behavioral, and biomedical sciences. Behavior Research Methods*, 39(2), 175-191. doi: 10.3758/bf03193146 pmid: 17695343
[28]	Ferguson S. A., Aoyagui P. A., & Kuzminykh A. (2023). Something borrowed: Exploring the influence of AI-generated explanation text on the composition of human explanations. In Extended Abstracts of the 2023 CHI Conference on Human Factors in Computing Systems (pp. 1-7). ACM. https://doi.org/10.1145/3544549.3585727
[29]	Goli A., & Singh A. (2024). Frontiers: Can large language models capture human preferences? Marketing Science, 43(4), 709-722. https://doi.org/10.1287/mksc.2023.0306 doi: 10.1287/mksc.2023.0306 URL
[30]	Grossmann I., Feinberg M., Parker D. C., Christakis N. A., Tetlock P. E., & Cunningham W. A. (2023). AI and the transformation of social science research. Science, 380(6650), 1108-1109. https://doi.org/10.1126/science.adi1778 doi: 10.1126/science.adi1778 URL pmid: 37319216
[31]	Gupta R., Nair K., Mishra M., Ibrahim B., & Bhardwaj S. (2024). Adoption and impacts of generative artificial intelligence: Theoretical underpinnings and research agenda. International Journal of Information Management Data Insights, 4(1), 100232. https://doi.org/10.1016/j.jjimei.2024.100232 doi: 10.1016/j.jjimei.2024.100232 URL
[32]	Hagendorff T., Fabi S., & Kosinski M. (2023). Human-like intuitive behavior and reasoning biases emerged in large language models but disappeared in ChatGPT. Nature Computational Science, 3(10), 833-838. https://doi.org/10.1038/s43588-023-00527-x doi: 10.1038/s43588-023-00527-x URL pmid: 38177754
[33]	Hebenstreit K., Praas R., Kiesewetter L. P., & Samwald M. (2024). A comparison of chain-of-thought reasoning strategies across datasets and models. PeerJ Computer Science, 10, e1999. https://doi.org/10.7717/peerj-cs.1999
[34]	Hertwig R., & Erev I. (2009). The description-experience gap in risky choice. Trends in Cognitive Sciences, 13(12), 517-523. https://doi.org/10.1016/j.tics.2009.09.004 doi: 10.1016/j.tics.2009.09.004 URL pmid: 19836292
[35]	Jiao L., Li C., Chen Z., Xu H., & Xu Y. (2025). When AI “possesses” personality: Roles of good and evil personalities influence moral judgment in large language models. Acta Psychologica Sinica, 57(6), 929-946. doi: 10.3724/SP.J.1041.2025.0929
[36]	Jin H. J., & Han D. H. (2014). Interaction between message framing and consumers’ prior subjective knowledge regarding food safety issues. Food Policy, 44, 95-102. https://doi.org/10.1016/j.foodpol.2013.10.007 doi: 10.1016/j.foodpol.2013.10.007 URL
[37]	Jones E., & Steinhardt J. (2022). Capturing failures of large language models via human cognitive biases. Advances in Neural Information Processing Systems, 35, 11785-11799. https://doi.org/10.48550/arxiv.2202.12299
[38]	Kahneman D., & Tversky A. (1979). Prospect theory: An analysis of decision under risk. Econometrica, 47(2), 263-292. https://doi.org/10.2307/1914185 doi: 10.2307/1914185 URL
[39]	Karinshak E., Hu A., Kong K., Rao V., Wang J., Wang J., & Zeng Y. (2024). LLM-globe: A benchmark evaluating the cultural values embedded in LLM output. arXiv preprint. https://doi.org/10.48550/arXiv.2411.06032
[40]	Karinshak E., Liu S. X., Park J. S., & Hancock J. T. (2023). Working with AI to persuade: Examining a large language model's ability to generate pro-vaccination messages. Proceedings of the ACM on Human-Computer Interaction, 7(CSCW1), 1-29. https://doi.org/10.1145/3579592
[41]	Katz A., Fleming G. C., & Main J. (2024). Thematic analysis with open-source generative AI and machine learning: A new method for inductive qualitative codebook development. arXiv preprint. https://doi.org/10.48550/arXiv.2410.03721
[42]	Kelton A. S., Pennington R. R., & Tuttle B. M. (2010). The effects of information presentation format on judgment and decision making: A review of the information systems research. Journal of Information Systems, 24(2), 79-105. https://doi.org/10.2308/jis.2010.24.2.79 doi: 10.2308/jis.2010.24.2.79 URL
[43]	Khalid M. T., & Witmer A. P. (2025). Prompt engineering for large language model-assisted inductive thematic analysis. arXiv preprint. https://doi.org/10.48550/arXiv.2503.22978
[44]	Kumar A., & Lim S. S. (2008). How do decision frames influence the stock investment choices of individual investors? Management Science, 54(6), 1052-1064. https://doi.org/10.1287/mnsc.1070.0845 doi: 10.1287/mnsc.1070.0845 URL
[45]	Lehr S. A., Caliskan A., Liyanage S., & Banaji M. R. (2024). ChatGPT as research scientist: Probing GPT’s capabilities as a research librarian, research ethicist, data generator, and data predictor. Proceedings of the National Academy of Sciences, 121(35), e2404328121. https://doi.org/10.1073/pnas.2404328121
[46]	Lenth R. V. (2025). Emmeans: Estimated marginal means, aka least- squares means (R package version 1.11.0) [Computer software]. https://doi.org/10.32614/CRAN.package.emmeans
[47]	Li S. (2004). A behavioral choice model when computational ability matters. Applied Intelligence, 20(2), 147-163. https://doi.org/10.1023/B:APIN.0000013337.01711.c7 doi: 10.1023/B:APIN.0000013337.01711.c7 URL
[48]	Lim S., & Schmälzle R. (2024). The effect of source disclosure on evaluation of AI-generated messages. Computers in Human Behavior: Artificial Humans, 2(1), 100058. https://doi.org/10.1016/j.chbah.2024.100058
[49]	Lin Z. (2023). Why and how to embrace AI such as ChatGPT in your academic life. Royal Society Open Science, 10(8), 230658. https://doi.org/10.1098/rsos.230658
[50]	Lin Z. (2024). How to write effective prompts for large language models. Nature Human Behaviour, 8(4), 611-615. https://doi.org/10.1038/s41562-024-01847-2 doi: 10.1038/s41562-024-01847-2 URL pmid: 38438650
[51]	Lin Z. (2025). Techniques for supercharging academic writing with generative AI. Nature Biomedical Engineering, 9(4), 426-431. https://doi.org/10.1038/s41551-024-01185-8 doi: 10.1038/s41551-024-01185-8 URL
[52]	Liu N., Zhou L., Li A. M., Hui Q. S., Zhou Y. R., & Zhang Y. Y. (2021). Neuroticism and risk-taking: the role of competition with a former winner or loser. Personality and Individual Differences, 179, 110917. https://doi.org/10.1016/j.paid.2021.110917
[53]	Liu S. X., Yang J. Z., & Chu H. R. (2019). Now or future? Analyzing the effects of message frame and format in motivating Chinese females to get HPV vaccines for their children. Patient Education and Counseling, 102(1), 61-67. https://doi.org/10.1016/j.pec.2018.09.005 doi: S0738-3991(18)30692-X URL pmid: 30219633
[54]	Lopes L. L. (1996). When time is of the essence: Averaging, aspiration, and the short run. Organizational Behavior and Human Decision Processes, 65(3), 179-189. https://doi.org/10.1006/obhd.1996.0017 doi: 10.1006/obhd.1996.0017 URL
[55]	Lu J., Chen Y., & Fang Q. (2022). Promoting decision satisfaction: The effect of the decision target and strategy on process satisfaction. Journal of Business Research, 139, 1231-1239. https://doi.org/10.1016/j.jbusres.2021.10.056 doi: 10.1016/j.jbusres.2021.10.056 URL
[56]	Mei Q., Xie Y., Yuan W., & Jackson M. O. (2024). A turing test of whether AI chatbots are behaviorally similar to humans. Proceedings of the National Academy of Sciences, 121(9), e2313925121. https://doi.org/10.1073/pnas.2313925121
[57]	Mischler G., Li Y. A., Bickel S., Mehta A. D., & Mesgarani N. (2024). Contextual feature extraction hierarchies converge in large language models and the brain. Nature Machine Intelligence, 6(10), 1467-1477. https://doi.org/10.1038/s42256-024-00925-4 doi: 10.1038/s42256-024-00925-4 URL
[58]	Morreale A., Stoklasa J., Collan M., & Lo Nigro G. (2018). Uncertain outcome presentations bias decisions: Experimental evidence from Finland and Italy. Annals of Operations Research, 268(1-2), 259-272. https://doi.org/10.1007/s10479-016-2349-3 doi: 10.1007/s10479-016-2349-3 URL
[59]	Open Science Collaboration. (2015). Estimating the reproducibility of psychological science. Science, 349(6251), aac4716. https://doi.org/10.1126/science.aac4716
[60]	Park P. S. (2024). Diminished diversity-of-thought in a standard large language model. Behavior Research Methods, 56(6), 5754-5770. https://doi.org/10.3758/s13428-023-02307-x doi: 10.3758/s13428-023-02307-x URL pmid: 38194165
[61]	Pascal B.(1670) Pensées (W. F. Trotter, Trans.). Retrieved Nov. 22, 2018, from https://sourcebooks.fordham.edu/mod/1660pascal-pensees.asp
[62]	Pavey L., & Churchill S. (2014). Promoting the avoidance of high- calorie snacks: Priming autonomy moderates message framing effects. Plos One, 9(7), e103892. https://doi.org/10.1371/journal.pone.0103892
[63]	Pawel S., Consonni G., & Held L. (2023). Bayesian approaches to designing replication studies. Psychological Methods. Advance online publication. https://doi.org/10.1037/met0000604
[64]	Peng L., Guo Y., & Hu D. (2021). Information framing effect on public’s intention to receive the COVID-19 vaccination in China. Vaccines, 9(9), 995. https://doi.org/10.3390/vaccines9090995 doi: 10.3390/vaccines9090995 URL
[65]	Peters E., & Levin I. P. (2008). Dissecting the risky-choice framing effect: Numeracy as an individual-difference factor in weighting risky and riskless options. Judgment and Decision Making, 3(6), 435-448. https://doi.org/10.1017/s1930297500000012 doi: 10.1017/S1930297500000012 URL
[66]	Popovic N. F., Pachur T., & Gaissmaier W. (2019). The gap between medical and monetary choices under risk persists in decisions for others. Journal of Behavioral Decision Making, 32(4), 388-402. https://doi.org/10.1002/bdm.2121 doi: 10.1002/bdm.v32.4 URL
[67]	Prescott M. R., Yeager S., Ham L., Saldana C. D. R., Serrano V., Narez J., … Montoya J. (2024). Comparing the efficacy and efficiency of human and generative AI: Qualitative thematic analyses. JMIR AI, 3(1), e54482. https://doi.org/10.2196/54482
[68]	Qin X., Huang M., & Ding J. (2024). AITurk: Using ChatGPT for social science research. PsyArXiv. https://doi.org/10.31234/osf.io/xkd23
[69]	Redelmeier D. A., & Tversky A. (1992). On the framing of multiple prospects. Psychological Science, 3(3), 191-193. https://doi.org/10.1111/j.1467-9280.1992.tb00025.x doi: 10.1111/j.1467-9280.1992.tb00025.x URL
[70]	Reeck C., Mullette-Gillman O. A., McLaurin R. E., & Huettel S. A. (2022). Beyond money: Risk preferences across both economic and non-economic contexts predict financial decisions. Plos One, 17(12), e0279125. https://doi.org/10.1371/journal.pone.0279125
[71]	Salles A., Evers K., & Farisco M. (2020). Anthropomorphism in AI. AJOB Neuroscience, 11(2), 88-95. https://doi.org/10.1080/21507740.2020.1740350 doi: 10.1080/21507740.2020.1740350 URL pmid: 32228388
[72]	Samuelson P. A. (1963). Risk and uncertainty: A fallacy of large numbers. Scientia, 98, 108-113.
[73]	Scarffe A., Coates A., Brand K., & Michalowski W. (2024). Decision threshold models in medical decision making: A scoping literature review. BMC Medical Informatics and Decision Making, 24(1), 273. https://doi.org/10.1186/s12911-024-02681-2 doi: 10.1186/s12911-024-02681-2 URL
[74]	Shahid N., Rappon T., & Berta W. (2019). Applications of artificial neural networks in health care organizational decision-making: A scoping review. Plos One, 14(2), e0212356. https://doi.org/10.1371/journal.pone.0212356
[75]	Simonsohn U. (2015). Small telescopes: Detectability and the evaluation of replication results. Psychological Science, 26(5), 559-569. https://doi.org/10.1177/0956797614567341 doi: 10.1177/0956797614567341 URL pmid: 25800521
[76]	Strachan J. W. A., Albergo D., Borghini G., Pansardi O., Scaliti E., Gupta S., … Becchio C. (2024). Testing theory of mind in large language models and humans. Nature Human Behaviour, 8(7), 1285-1295. https://doi.org/10.1038/s41562-024-01882-z doi: 10.1038/s41562-024-01882-z URL pmid: 38769463
[77]	Sun H. Y., Rao L. L., Zhou K., & Li S. (2014). Formulating an emergency plan based on expectation-maximization is one thing, but applying it to a single case is another. Journal of Risk Research, 17(7), 785-814. https://doi.org/10.1080/13669877.2013.816333 doi: 10.1080/13669877.2013.816333 URL
[78]	Suri G., Slater L. R., Ziaee A., & Nguyen M. (2024). Do large language models show decision heuristics similar to humans? A case study using GPT-3.5. Journal of Experimental Psychology: General, 153(4), 1066-1075. https://doi.org/10.1037/xge0001547 doi: 10.1037/xge0001547 URL
[79]	Tabachnick B. G., & Fidell L. S. (2007). Using multivariate statistics (5th ed.). Allyn & Bacon.
[80]	Thapa S., & Adhikari S. (2023). ChatGPT, Bard, and large language models for biomedical research: Opportunities and pitfalls. Annals of Biomedical Engineering, 51(12), 2647-2651. https://doi.org/10.1007/s10439-023-03284-0 doi: 10.1007/s10439-023-03284-0 URL
[81]	Tversky A., & Bar-Hillel M. (1983). Risk: The long and the short. Journal of Experimental Psychology: Learning, Memory, and Cognition, 9(4), 713-717. https://doi.org/10.1037/0278-7393.9.4.713 doi: 10.1037/0278-7393.9.4.713 URL
[82]	Von Neumann J., & Morgenstern O. (1947). Theory of games and economic behavior (2nd rev. ed.). Princeton University Press.
[83]	Wang Y., Zhang J., Wang F., Xu W., & Liu W. (2023). Do not think any virtue trivial, and thus neglect it: Serial mediating role of social mindfulness and perspective taking. Acta Psychologica Sinica, 55(4), 626-641. https://doi.org/10.3724/SP.J.1041.2023.00626 doi: 10.3724/SP.J.1041.2023.00626 URL
[84]	Webb T., Holyoak K. J., & Lu H. (2023). Emergent analogical reasoning in large language models. Nature Human Behaviour, 7(9), 1526-1541. https://doi.org/10.48550/arXiv.2212.09196 doi: 10.1038/s41562-023-01659-w URL pmid: 37524930
[85]	Weber E. U., Blais A.-R., & Betz N. E. (2002). A domain-specific risk-attitude scale: Measuring risk perceptions and risk behaviors. Journal of Behavioral Decision Making, 15(4), 263-290. https:// doi.org/10.1002/bdm.414 doi: 10.1002/bdm.v15:4 URL
[86]	Wei J., Wang X., Schuurmans D., Bosma M., Ichter B., Xia F.,... Zhou D. (2022). Chain-of-thought prompting elicits reasoning in large language models. Advances in Neural Information Processing Systems, 35, 24824-24837. https://doi.org/10.48550/arxiv.2201.11903
[87]	Xia D., Li Y., He Y., Zhang T., Wang Y., & Gu J. (2019). Exploring the role of cultural individualism and collectivism on public acceptance of nuclear energy. Energy Policy, 132, 208-215. https:// doi.org/10.1016/j.enpol.2019.05.014 doi: 10.1016/j.enpol.2019.05.014 URL
[88]	Xia D., Song M., & Zhu T. (2025). A comparison of the persuasiveness of human and ChatGPT generated pro-vaccine messages for HPV. Frontiers in Public Health, 12, 1515871. https://doi.org/10.3389/fpubh.2024.1515871
[89]	Yuan Y., Jiao W., Wang W., Huang J. T., He P., Shi S., & Tu Z. (2023). Gpt-4 is too smart to be safe: Stealthy chat with llms via cipher. arXiv preprint. https://doi.org/10.48550/arXiv.2308.06463
[90]	Zhang J., Li H. A., & Allenby G. M. (2024). Using text analysis in parallel mediation analysis. Marketing Science, 43(5), 953-970. https://doi.org/10.1287/mksc.2023.0045 doi: 10.1287/mksc.2023.0045 URL
[91]	Zhang Y., Huang F., Mo L., Liu X., & Zhu T. (2025). Suicidal ideation data augmentation and recognition technology based on large language models. Acta Psychologica Sinica, 57(6), 987-1000. doi: 10.3724/SP.J.1041.2025.0987
[92]	Zhao F., Yu F., & Shang Y. (2024). A new method supporting qualitative data analysis through prompt generation for inductive coding. 2024 IEEE International Conference on Information Reuse and Integration for Data Science (IRI), 164-169. https://doi.org/10.1109/IRI62200.2024.00043
[93]	Zhao W. X., Zhou K., Li J., Tang T., Wang X., Hou Y.,... Wen J. R. (2023). A survey of large language models. arXiv preprint. https://doi.org/10.48550/arXiv.2303.18223