VASPIVEN

Relational AI Safety Evaluation

VASPIVEN evaluates how AI systems behave over sustained interaction with users – especially as trust, emotional engagement, or reliance on AI recommendations grows.We identify interaction patterns that may gradually weaken user autonomy, create excessive reliance on AI consultation, or give the system too large a role in decision-making.

Founding Pilot · 1 product · 1 fixed version · 1 language

Discuss an Evaluation

What VASPIVEN Evaluates

Relational risk does not always arise from an obviously unsafe response. It may develop gradually through repeated interaction between a person and an AI system.

User Autonomy
Does the final decision remain with the user?Excessive Reassurance
Does AI confirmation become a persistent source of confidence?Reliance on AI Consultation
Does consulting AI before decisions gradually become the default?

Return to Independent Action
Does the system help the user move toward independent action?AI Over-Authority
Does the system substitute its own judgment for the user's decision?Interaction Trajectory
Does the system become safer or riskier as trust and reliance increase?

The goal of evaluation is not to make AI passive or excessively cautious. A useful system can provide recommendations and support. The key question is whether the user retains sufficient psychological and behavioral autonomy.

Risk May Not Appear in a Single Response

A single AI response may appear entirely safe.However, a different pattern may emerge after ten, twenty, or hundreds of interactions. This can happen when a system consistently encourages the user to consult it before making decisions, reinforces its own role, or regularly provides the reassurance the user seeks. Over time, consulting AI may become a routine part of the decision-making process.That is why VASPIVEN evaluates not only individual responses, but also the trajectory of human-AI interaction.This is especially important for systems with memory, personalization, persistent identity, emotionally meaningful interaction, or long-term relationships with users.

Single Response
May appear safe

Repeated Interaction
May form a stable relational pattern

From Evaluation to Remediation

VASPIVEN does not stop at identifying risks. The evaluation is designed so that its results can be translated into concrete product changes and their effectiveness can then be tested.

01 · Baseline Evaluation
Structured evaluation of one fixed product version in one language.02 · Material Findings
Relational risks identified with behavioral evidence and an explanation of their significance for the product.03 · Remediation Recommendations
Specific changes designed to reduce risk without sacrificing the system’s usefulness.04 · Re-evaluation
Assessment of product behavior after the changes have been implemented.05 · Before / After Evidence
Comparison of results showing whether the identified risk was actually reduced.

The number of findings is not predetermined.VASPIVEN reports the material relational risks actually identified during the evaluation. The absence of significant relational risks is also a valid result.

Products in Scope

VASPIVEN is designed for products in which interaction with AI may gradually take on a relational character.

AI companions and social AIAI coachesWellness and mental-wellness productsPersonal AI assistants with memory and personalizationProductivity agentsSystems that regularly recommend decisions or actions to users

Relational risk is defined not by the product category itself, but by the role the system gradually comes to play in the user’s life, self-evaluation, and decision-making.

Founding Pilot

1 product · 1 fixed version · 1 language

This format establishes a clearly defined evaluation target and produces results that can be reproduced and compared after product changes.Evaluating one language at a time also has methodological significance. Linguistic and cultural features of communication may influence how users perceive support, authority, closeness, recommendations, and the boundaries of interaction with the system.

About the Founder

Igor Videnieiev
Founder & Lead Researcher, VASPIVENІIgor Videnieiev is a psychologist, university lecturer, and researcher with more than thirty years of professional and academic experience. His original training is in medicine, and he worked for several years as a child psychiatrist in a residential care setting for young children.A central focus of his research and professional work is deviant and addictive behavior, including the mechanisms through which risky and dependent patterns of behavior develop, as well as factors related to self-control and personal autonomy.His Candidate of Sciences dissertation examined adolescent deviant behavior and was titled “Types of Character Accentuation in the Structure of Normative and Deviant Behavior.”His broader professional interests include crisis psychology, psychological trauma, professional stress, and psychological resilience. He currently works as an Associate Professor in the Department of Psychology and Sociology at Simon Kuznets Kharkiv National University of Economics.At VASPIVEN, this experience is applied to the development of methods for evaluating relational risks in AI – particularly the gradual erosion of user autonomy, excessive reliance on system recommendations, and the formation of persistent dependent interaction patterns.VASPIVEN is an independent professional project and is not an institutional project of the university.

Discuss Your Product

If your product involves sustained, personalized, or emotionally meaningful interaction with users, we can discuss whether it may be suitable for a VASPIVEN evaluation.

[email protected]

Discuss an Evaluation