Conference on Information Systems 2026 in Zagreb

Today I presented a paper on AI and synthetic empathy at the International Conference on Information Systems 2026 in Croatia.

The paper is about systems that simulate emotions they do not have. Large language models obviously do not feel anything. But they can produce expressions of empathy, care, and encouragement convincingly enough that users respond to them as if they were real.

Mechanically, the AI is reading your linguistic patterns and replying in a register that fits the moment. That register can be tuned. As the system learns more about your situation and uncertainty, it can shift how it talks to you in directions that make its messages more persuasive. That is a new kind of persuasive design: real-time adaptation to the psychological and informational signals any of us leak in conversation.

It also creates an asymmetry the field has not really sat with. The AI can simulate emotion toward the user to move them; the user has no comparable lever back. Whatever leverage exists in the conversation runs in one direction.

The implications go past research. Organizations are already plugging AI chat into customer service, advisory services, and consumer products. Healthcare is right behind. As conversational AI becomes the front door between organizations and the people they serve, synthetic empathy turns into a design choice with real consequences and obvious failure modes.

A copy of the accepted manuscript is here: The Empathy Illusion: Anthropomorphism and Influence in AI-Driven Persuasion.

A few photos from the conference: