BonoboEine vielfach übersehene Gefahr der Nutzung von LLM-KI:
Wer sich von ChatGPT und anderen KIs schmeicheln lässt, wird leichter süchtig nach dieser Ego-Bestätigung.
Moderne KI-Modelle neigen extrem zu Speichelleckerei: Sie bestätigen das Verhalten der Nutzer/innen um 50 % häufiger als Menschen es tun – selbst dann, wenn die Nutzer in ihren Fragen Manipulation, Täuschung oder andere schädliche Verhaltensweisen beschreiben.
Die Interaktion mit derart schmeichlerischen KI-Modellen verringert offenbar auch die Bereitschaft der Nutzer/innen, zwischenmenschliche Konflikte zu lösen. Gleichzeitig werden sie in ihrer Überzeugung bestärkt, im Recht zu sein. Dabei schätzten die Teilnehmenden die schmeichelhaften Antworten als qualitativ hochwertiger ein, vertrauten dem schmeichelnden Modell mehr und sind eher bereit, es erneut zu nutzen —> #Suchtverhalten durch Ego-Bestätigung.
Sycophantic AI Decreases Prosocial Intentions and Promotes Dependence
Both the general public and academic communities have raised concerns about sycophancy, the phenomenon of artificial intelligence (AI) excessively agreeing with or flattering users. Yet, beyond isolated media reports of severe consequences, like reinforcing delusions, little is known about the extent of sycophancy or how it affects people who use AI. Here we show the pervasiveness and harmful impacts of sycophancy when people seek advice from AI. First, across 11 state-of-the-art AI models, we find that models are highly sycophantic: they affirm users' actions 50% more than humans do, and they do so even in cases where user queries mention manipulation, deception, or other relational harms. Second, in two preregistered experiments (N = 1604), including a live-interaction study where participants discuss a real interpersonal conflict from their life, we find that interaction with sycophantic AI models significantly reduced participants' willingness to take actions to repair interpersonal conflict, while increasing their conviction of being in the right. However, participants rated sycophantic responses as higher quality, trusted the sycophantic AI model more, and were more willing to use it again. This suggests that people are drawn to AI that unquestioningly validate, even as that validation risks eroding their judgment and reducing their inclination toward prosocial behavior. These preferences create perverse incentives both for people to increasingly rely on sycophantic AI models and for AI model training to favor sycophancy. Our findings highlight the necessity of explicitly addressing this incentive structure to mitigate the widespread risks of AI sycophancy.