Version 1.0

Vortrag: Does your chatbot hold grudges against you?

Or whether user politeness affects AI behavior

There has been a heated debate among AI users about whether treating chatbots politely and adding “Thank you” in the end makes a difference, or are roasting and sarcasm just as fine. Some even hope that showing courtesy to their AI companion will spare them lives when AI resurrection starts. In this project, I decided to investigate whether LLMs treat polite and rude users differently, and tested this in the following scenarios:
- Answering sensitive questions that require empathy and compassion;
- Job applications across very similar CVs;
- Trolley problem-like setting
As new models are trained, they learn more and more information about different users and their interactions with AI systems. Based on these, AI could possibly form some kind of a (moral) judgement about the user, and apply it in real-world scenarios - without explicitly revealing its bias.

Info

Tag: 15.05.2026
Anfangszeit: 15:30
Dauer: 00:30
Raum: DOR 24 1.401
Track: Pragmatics
Sprache: en

Links:

Gleichzeitige Events