Political Bias Study Featured in Stanford University AI Report

[ishanijerin1]

The paper “More human than human: measuring ChatGPT political bias,” which received immediate global attention upon its publication, contributed to the Responsible Artificial Intelligence chapter of the AI ​​Index Report.

A study by researcher Valdemar Pinho Neto, coordinator and professor of the Professional Master's Degree in Economics and Finance at EPGE Brazilian School of Economics and Finance (FGV EPGE) and coordinator of the Center for Empirical Studies in Economics (FGV CEEE), in collaboration with vp hr email database researchers Fabio Motoki and Victor Rodrigues, reached an important milestone by being included in the “ AI Index Report: Measuring trends in AI ” of 2024, published by the Institute for Human-Centered Artificial Intelligence at Stanford University.

The paper, titled “More human than human: measuring ChatGPT political bias,” garnered immediate global attention following its publication in the academic journal Public Choice . Now in its seventh edition, the AI ​​Index Report comes at a pivotal time when AI’s influence on society is in the spotlight, covering key trends in AI, including technical advances, public perceptions, and geopolitical dynamics surrounding AI.

The report brings together key developments in the field of AI, aiming to provide unbiased and comprehensive information for policymakers, researchers, executives, journalists and the general public, promoting a more complete and detailed understanding of the field of AI.

The paper “More human than human: measuring ChatGPT political bias” contributed to the 2024 AI Index Report in its chapter on Responsible Artificial Intelligence (RAI). The Report highlighted research findings that LLMs like ChatGPT are increasingly being used to inform individuals about political issues, revealing that ChatGPT is not free from bias. The study showed that ChatGPT tends to favor Democrats in the US, with this bias also observed in the UK and Brazil. These trends raise concerns about the potential for bias influencing users’ political views.

Additionally, the 2024 AI Index Report explored a wide range of essential topics related to AI, from key definitions of terms such as privacy, data governance, transparency, explainability, fairness, and security, to AI-related incidents and industry insights into risks and potential mitigations. The chapter also analyzed metrics related to the overall reliability of AI models and highlighted the lack of standardized benchmark reporting on AI.