Assessing Political Bias and Value Misalignment in Generative Artificial Intelligence

Fabio Yoshio Suguri Motoki; Valdemar Pinho Neto; Victor Rodrigues

Assessing Political Bias and Value Misalignment in Generative Artificial Intelligence

Fabio Yoshio Suguri Motoki, Valdemar Pinho Neto, Victor Rodrigues

Research output: Working paper › Preprint

1 Downloads (Pure)

Abstract

Our analysis reveals a concerning misalignment of values between ChatGPT and the average American. We also show that ChatGPT displays political leanings when generating text and images, but the degree and direction of skew depend on the theme. Notably, ChatGPT repeatedly refused to generate content representing certain mainstream perspectives, citing concerns over misinformation and bias. As generative AI systems like ChatGPT become ubiquitous, such misalignment with societal norms poses risks of distorting public discourse. Without proper safeguards, these systems threaten to exacerbate societal divides and depart from principles that underpin free societies.

Original language	English
Publisher	SSRN
Publication status	Published - 27 Mar 2024

Keywords

Generative AI
Societal values
Large language models
Multimodal
AI governance

Access to Document

https://ssrn.com/abstract=4773936Licence: CC BY-NC-ND

1 Article

Assessing political bias and value misalignment in generative artificial intelligence
Motoki, F. Y. S., Pinho Neto, V. & Rodrigues, V., 4 Feb 2025, (E-pub ahead of print) In: Journal of Economic Behavior & Organization. 106904.
Research output: Contribution to journal › Article › peer-review

Open Access

1 Invited talk

Transformando dados em decisões: Inteligência Artificial aplicada à Economia e Finanças
Fabio Motoki (Invited speaker) & Valdemar Pinho Neto (Chair)
27 Sept 2024
Activity: Participating in or organising an event › Invited talk

Cite this

@techreport{90967f778cdf42ecb4cc59b303cc12fb,

title = "Assessing Political Bias and Value Misalignment in Generative Artificial Intelligence",

abstract = "Our analysis reveals a concerning misalignment of values between ChatGPT and the average American. We also show that ChatGPT displays political leanings when generating text and images, but the degree and direction of skew depend on the theme. Notably, ChatGPT repeatedly refused to generate content representing certain mainstream perspectives, citing concerns over misinformation and bias. As generative AI systems like ChatGPT become ubiquitous, such misalignment with societal norms poses risks of distorting public discourse. Without proper safeguards, these systems threaten to exacerbate societal divides and depart from principles that underpin free societies.",

keywords = "Generative AI, Societal values, Large language models, Multimodal, AI governance",

author = "Motoki, {Fabio Yoshio Suguri} and {Pinho Neto}, Valdemar and Victor Rodrigues",

year = "2024",

month = mar,

day = "27",

language = "English",

publisher = "SSRN",

type = "WorkingPaper",

institution = "SSRN",

}

TY - UNPB

T1 - Assessing Political Bias and Value Misalignment in Generative Artificial Intelligence

AU - Motoki, Fabio Yoshio Suguri

AU - Pinho Neto, Valdemar

AU - Rodrigues, Victor

PY - 2024/3/27

Y1 - 2024/3/27

N2 - Our analysis reveals a concerning misalignment of values between ChatGPT and the average American. We also show that ChatGPT displays political leanings when generating text and images, but the degree and direction of skew depend on the theme. Notably, ChatGPT repeatedly refused to generate content representing certain mainstream perspectives, citing concerns over misinformation and bias. As generative AI systems like ChatGPT become ubiquitous, such misalignment with societal norms poses risks of distorting public discourse. Without proper safeguards, these systems threaten to exacerbate societal divides and depart from principles that underpin free societies.

AB - Our analysis reveals a concerning misalignment of values between ChatGPT and the average American. We also show that ChatGPT displays political leanings when generating text and images, but the degree and direction of skew depend on the theme. Notably, ChatGPT repeatedly refused to generate content representing certain mainstream perspectives, citing concerns over misinformation and bias. As generative AI systems like ChatGPT become ubiquitous, such misalignment with societal norms poses risks of distorting public discourse. Without proper safeguards, these systems threaten to exacerbate societal divides and depart from principles that underpin free societies.

KW - Generative AI

KW - Societal values

KW - Large language models

KW - Multimodal

KW - AI governance

M3 - Preprint

BT - Assessing Political Bias and Value Misalignment in Generative Artificial Intelligence

PB - SSRN

ER -

Assessing Political Bias and Value Misalignment in Generative Artificial Intelligence

Abstract

Keywords

Access to Document

Research output

Assessing political bias and value misalignment in generative artificial intelligence

Activities

Transformando dados em decisões: Inteligência Artificial aplicada à Economia e Finanças

Cite this