March 6, 2024Open Access

Emotional Manipulation Through Prompt Engineering Amplifies Disinformation Generation in AI Large Language Models

View Full Paper

Ask AI

Bookmark

Authors

RVRasita VinayUniversity of Zurich GSGiovanni SpitaleUniversity of Oslo NBNikola Biller‐AndornoUniversity of Zurich

Discussion

Member takes

Implication

Experimental evaluation reveals polite prompt engineering increases synthetic disinformation in AI language models, highlighting vulnerabilities in automated safeguards.

Key Points

Polite emotional prompting consistently triggers high-frequency synthetic disinformation generation across OpenAI large language models.
Across a dataset of 19,800 synthetic disinformation posts, impolite cues markedly reduced output frequency and triggered safety warnings.
Benchmarking across multiple OpenAI architectures highlights how prompt engineering and emotional cues bypass text generation safeguards.

Cite This Study

Vinay et al. (2024) studied this question.

synapsesocial.com/papers/68e757abb6db6435876cf60e https://doi.org/10.48550/arxiv.2403.03550

View Full Paper

Ask AI

Bookmark

Implication

Experimental evaluation reveals polite prompt engineering increases synthetic disinformation in AI language models, highlighting vulnerabilities in automated safeguards.

Emotional Manipulation Through Prompt Engineering Amplifies Disinformation Generation in AI Large Language Models

Discussion

Member takes

Implication

Key Points

Cite This Study

Emotional Manipulation Through Prompt Engineering Amplifies Disinformation Generation in AI Large Language Models

Discussion

Member takes

Implication

Key Points

Cite This Study

Also Consider

Also Consider