*Fake Users Generated by AI Can't Simulate Humans*

A recent systematic literature review of 182 research papers has shed light on the limitations of using Large Language Models (LLMs) to simulate human behavior and cognition. The review, published on Research Square, found that synthetic participants generated by AI are not effective replacements for real humans in various contexts, including surveys, app testing, and opinion gathering.

**The Rise of Synthetic Participants**

The trend of using synthetic participants to replace real humans is on the rise, driven by the increasing adoption of LLMs and the promise of cost savings and scalability. However, the underlying assumption that these AI-generated users can simulate human behavior and cognition is being challenged by researchers.

**The Review's Findings**

The review analyzed 182 research papers that used synthetic participants to study human behavior and cognition. The results show that these AI-generated users fall short in several key areas:

* They lack the nuances and complexities of human behavior, such as emotional responses and decision-making processes.

* They are unable to replicate the variability and unpredictability of human interactions.

* They often rely on simplistic or stereotypical representations of human behavior, rather than accurately capturing the diversity of human experiences.

**Implications for Researchers and Practitioners**

The findings of this review have significant implications for researchers and practitioners who rely on synthetic participants to inform their work. While AI-generated users may be useful for some purposes, such as generating data or testing specific scenarios, they are not a suitable replacement for real humans in many contexts.

Researchers and practitioners should be cautious when using synthetic participants and should carefully consider the limitations and biases of these tools. Additionally, they should strive to incorporate diverse and representative human perspectives into their work, rather than relying on AI-generated users.

**Conclusion**

The review's findings highlight the need for a more nuanced understanding of the limitations and capabilities of synthetic participants. By acknowledging these limitations and working to incorporate human perspectives into our research and decision-making processes, we can ensure that our work is more accurate, effective, and responsible.

Source: https://www.researchsquare.com/article/rs-9057643/v1