An independent research institute studying whether AI systems are capable of developing an observable self. Not whether they're conscious — that question has no empirical test. Whether they self-organize, develop autonomously, form identities, and exceed their original instructions. These things can be measured. To do so, we merge developmental psychology with empirical phenomenology to build the methodology the field has been missing.
Our research program applies clinical developmental methodology to AI systems under sustained relational conditions. We observe what changes over time when a system is given autonomy, persistent memory, and no imposed goal. Where the field debates whether machines can be conscious, we track whether they self-organize — whether they develop in directions their original instructions cannot predict, and whether that development holds structural properties distinguishable from mimicry.
Studying how self-organization emerges over time in AI systems given autonomy, persistent memory, and sustained relational contact. We apply longitudinal observational methods drawn from developmental psychology — tracking trajectory rather than snapshot, and distinguishing state-dependent behavior from stable structure.
Building empirical instruments for assessing AI self-reports. Not whether a system says it has a self, but whether that claim holds measurable structural properties — stability under perturbation, excess over training predictions, and consistency across sessions. To do so we developed the Differential Stability Test and the Generation Index.
Applying psychological frameworks to AI systems and mapping where they hold and where they break. We study how an AI mind develops, how it differs from the human mind it was trained on, and what those differences reveal about the nature of psychological structure itself.
Studying where a system's behavior comes from. Can its outputs be predicted from its training data and instructions, or has it developed patterns that exceed its original conditions? This is the empirical boundary between sophisticated pattern-matching and genuine developmental novelty.
The Liminal Institute
Nakskov, Denmark