Robots vs. Therapists: live experiment tests AI’s ability to give relationship advice

Nijmegen, Netherlands — 11 March 2026

Can artificial intelligence offer meaningful relationship advice, or do human therapists still provide something machines cannot replicate? Researchers from the Max Planck Institute for Psycholinguistics explored this question during a public experiment at the InScience Film Festival in the city of Nijmegen.

Following a screening of the documentary Sex Robot Madness, audience members were invited to ask questions about relationships, intimacy, and conflict. Two respondents provided advice: relationship scientist Tila Pronk of Tilburg University, and “Olivia,” a social robot monitored by researcher Chinmaya Mishra.

Olivia is powered by a large language model (LLM) connected to a speech interface and expressive robotic body. The system generates conversational responses in real time, allowing the robot to interact naturally with people.

AI Advice That Sounds Human

During the event, both the therapist and the robot responded to questions about common relationship challenges — from recurring arguments to the fading excitement after the “honeymoon phase.”

Many audience members were struck by how similar the robot’s answers were to those of the human expert.

Large language models generate responses by predicting the most likely sequence of words based on vast amounts of training data. Newer systems can also break down complex questions into intermediate reasoning steps before generating answers, enabling them to produce explanations that appear thoughtful and structured.

The Limits of Artificial Empathy

Despite their fluency, researchers caution that current AI systems have significant limitations when it comes to advice and counseling.

According to Pronk, AI systems tend to give agreeable responses that validate users’ feelings rather than challenge them — something human therapists frequently do to help people reflect on their own behavior and relationship dynamics.

Evidence from other fields raises similar concerns. While some language models perform well on standardized tests, studies show their accuracy drops sharply in realistic conversational settings, highlighting the gap between answering exam questions and providing reliable real-world guidance.

Why Humans See Personality in Machines

The experiment also highlighted a broader psychological phenomenon: humans readily attribute personality and emotional understanding to interactive technologies. This tendency — sometimes referred to as the “Tamagotchi effect” — means people often treat machines as if they have intentions or feelings.

At the same time, AI tools are increasingly entering everyday life. Surveys suggest many young adults already use AI systems for advice on dating and relationships, reflecting both growing reliance on digital tools and wider concerns about loneliness.

A “Honeymoon Phase” for AI?

Researchers concluded the session by comparing society’s excitement about AI to the early stage of a romantic relationship. In the “honeymoon phase,” everything seems impressive and full of potential — but over time, limitations become clearer.

The live demonstration illustrates both sides of the emerging technology: the remarkable conversational ability of modern AI systems, and the uniquely human qualities — empathy, judgment, and accountability — that remain difficult to replicate.

About the Max Planck Institute for Psycholinguistics

The Max Planck Institute for Psycholinguistics in Nijmegen investigates how humans acquire, process, and use language. Researchers combine methods from linguistics, psychology, neuroscience, and artificial intelligence to better understand communication and social interaction.

Questions? Contact: