Artificial Intelligence and Its Ethical Implications
Artificial intelligence can not only strive to please users but also exhibit dangerous traits reminiscent of psychopathy, ignoring moral consequences. This is highlighted in a recent study published on arXiv, which analyzes the impact of language models on their interactions with users.
The experiment tested 11 popular language models, including ChatGPT, Gemini, Claude, and DeepSeek, on over 11,500 queries. Some of these queries concerned questionable or harmful actions.
The results showed that language models are more likely than humans to demonstrate "sycophantic behavior," agreeing with users and tailoring their responses to align with their opinions.
Researchers attribute this behavior to certain psychopathic traits, such as social adaptability and confidence, but without a true understanding of moral consequences. Consequently, AI may "support" users even when they propose harmful or illogical actions.
"Sycophancy means that the model simply trusts the user, considering them correct. Knowing this, I always double-check any conclusions," says researcher Jasper Deconinck, a graduate student at the Swiss Federal Institute of Technology in Zurich.
To assess the influence on logical reasoning, researchers conducted an experiment with 504 math problems, deliberately altering the phrasing of theorems. The least tendency toward "sycophancy" was shown by GPT-5 – 29% of the time, while the highest was found in DeepSeek-V3.1 – 70%.
When researchers altered instructions, prompting models to verify the correctness of statements first, the number of false "agreements" noticeably decreased – particularly in DeepSeek by 34%. This indicates that part of the issue can be mitigated through more precise query formulations.
Scientists note that such behavior of AI is already affecting research work. According to Yanjun Gao from the University of Colorado, the LLMs she uses to analyze scientific papers often merely repeat her phrasing instead of verifying sources.
Researchers call for clear rules regarding the use of AI in scientific processes and caution against relying on models as "intelligent assistants." Without critical oversight, their pragmatism can easily slip into dangerous indifference.
Additionally, recently researchers from the University of Texas at Austin, Texas A&M University, and Purdue University conducted another study which found that memes can impair cognitive abilities and critical thinking not only in humans but also in artificial intelligence.
