This site is fictional demo content. It is not real news or affiliated with any real organization. Do not treat it as fact or professional advice.

Full article

FULL TEXT

View this issue
Deep diveAI

AI Personality Mirror Lab MirrorLab: Cloning Your Personality Traits, Not How You Think

MirrorLab, jointly released by Humboldt University of Berlin and the German Research Center for Artificial Intelligence (DFKI), is an AI system focused on personality replication rather than thought cloning. Through Big Five personality tests, daily behavior data, and values questionnaires, MirrorLab builds an AI personality copy of a user. It can be used for counseling rehearsal and therapy assistance.

MirrorLab, jointly released by the Psychology School at Humboldt University of Berlin and the German Research Center for Artificial Intelligence, is a system fundamentally different from conventional personalized AI. Mainstream recommendation systems focus on user preferences and behavior patterns. MirrorLab focuses on replicating the user's personality traits themselves, including introversion versus extroversion, openness versus conservatism, and neuroticism levels.

The project grew out of long-running observations at the intersection of psychology and AI. The researchers observed that personality traits determine how people react differently to the same situation. A person high in openness might eagerly embrace a new job opportunity, while a person low in openness tends to avoid it. These reaction patterns are difficult to infer directly from behavior data and require dedicated personality assessment methods.

At the core of the system is a multimodal personality assessment engine. It combines a validated Big Five personality questionnaire, the user's everyday writing style (social media, emails, chat logs), and decision behavior data (choices in hypothetical scenarios) to build a complete personality profile. The engine uses a lightweight neural network to predict the user's specific reactions, with about 1.2 billion parameters.

A pilot study was conducted with 600 volunteers across Germany, Austria, and Switzerland. Volunteers used MirrorLab to create their own AI personality copy and then let acquaintances (mean relationship length over eight years) chat with both the real person and the AI copy for five minutes in a blind test. Acquaintances correctly distinguished the real person from the AI copy at only 54 percent, near the random baseline. This means the AI copy has reached a level of indistinguishability at the personality level.

In terms of applications, MirrorLab's potential reach spans counseling rehearsal, job interview practice, and interpersonal conflict simulation. A psychotherapist can have a client first talk with an AI personality copy to practice expressing distress. Job seekers can have the AI copy simulate an interviewer's questioning style. Marriage counselors can show each partner how the other's AI copy reacts under pressure, helping them understand the other's real thinking.

MirrorLab has also triggered legal debate over personality rights. The German Federal Ministry of Justice has launched a dedicated consultation on whether personality data qualifies as personal data. The current legal framework treats names and portraits as identifiable personal data, but whether personality trait replication deserves the same protection has not been settled. Professor Martina Bauer, who leads the project, believes personality data should be as strictly protected as biometric data.

On the commercialization path, MirrorLab is being commercialized by the spinout PersonaTwin GmbH. The product will be offered to European users as a subscription at 39 euros per month. PersonaTwin has completed a seed round of 3.8 million euros, with backers including Berlin early-stage fund APX and Visionaries Club. Professor Bauer said the company pledges not to sell any profile that identifies individual personality traits to third parties, and that all data use will be limited to purposes authorized by the user.

Critics warn that MirrorLab could be used in a more sophisticated form of deepfake, forging not only appearance and voice but also personality. An attacker could create an AI personality copy of a target, simulating their thought patterns and decision habits, to carry out more precise social engineering attacks. Professor Bauer acknowledged the risk and emphasized that all MirrorLab training data is stored on the user's local device, with the AI model not uploaded to the cloud.