Testable predictions from §IX — falsifiable within specified timeframes and conditions.
01The Void Framework's value as a scientific instrument depends on its capacity to generate falsifiable predictions that discriminate between its account of AI companion harm and competing explanations rooted in individ…
02The predictions are ordered by temporal proximity, with the first checkable within twelve months of publication and the remainder structured around regulatory implementation timelines, longitudinal cohort availability…
03Prediction 1: EU AI Act Article 50 disclosure mandates will reduce new-user attachment formation but not existing-user dependency.
04Article 50 of the EU AI Act, enforceable from August 2, 2025, requires that users be informed when they are interacting with an AI system.
05Falsified if: Attachment language (as operationalized by Laestadius et al., 2022, Journal of Medical Internet Research, who found 35 percent of Replika reviews used attachment terminology) declines by more than 15 per…
06Prediction 2: Platforms scoring above Void Index 8 will generate measurably higher rates of user-reported emotional distress upon service discontinuation than platforms scoring below 5.