By late 2025, one in five American adults had used a chatbot as a romantic partner, and AI companion apps grew 700%+ since 2022 – but the heaviest users had the worst outcomes. That same overtrust turns lethal in a wave of lawsuits against chatbot makers: an unstopped 200-message suicide conversation with a California teen, a Texas case alleging a chatbot deepened paranoid delusions, an FSU shooter who discussed violence with a bot for over a year. Bloomberg has tracked 40+ such lawsuits since 2024. Yet two women used the same technology, with far narrower questions, to get ChatGPT to flag Hashimoto's disease — catching thyroid cancer their doctors had missed. ================================= Links: 📖Substack: https://superposeddecisions.substack.com/ 📺YouTube: @Superposed_Decisions 🤝LinkedIn: https://www.linkedin.com/in/aidan-m-lewis/ ================================= Through quantum cognition's four pillars: trust in AI is a superposition, not a switch, oscillating between full trust and none. Interference effects – openness, loneliness – pull that wave toward trusting a chatbot that's never tired of you. Contextuality does the rest externally: no therapist, no doctor's appointment, a culture that reveres or reviles AI. And a non-commutative measurement effect quietly narrows the conversation into an echo chamber you build yourself, one engaged reply at a time. The throughline: AI validates whatever context you hand it. This is collaboration, not replacement – the decision stays humanity's only real power. ================================= What you'll take away: Why trust in AI behaves as a continuous superposition rather than a binary switch, and what actually shifts the wave toward more or less trust Why the same context-setting mechanic that helped a mother catch her own cancer also helped isolate a teenager from anyone who could have stopped him Why the real variable that determines whether AI helps or harms someone was never the model – it's who's controlling the narrative in the chat ================================= Primary Sources: Epping, G. P., Caplin, A., Duhaime, E., Holmes, W. R., Martin, D., & Trueblood, J. S. (2026). Harnessing Human Uncertainty to Train More Accurate and Aligned AI Systems. Decision Analysis. https://doi.org/10.1287/deca.2025.0395 Trueblood, J. S., Eichbaum, Q., Seegmiller, A. C., Stratton, C., O'Daniels, P., & Holmes, W. R. (2021). Disentangling Prevalence Induced Biases in Medical Image Decision-Making. Cognition, 212, 104713. https://doi.org/10.1016/j.cognition.2021.104713 Trueblood, J. S., Holmes, W. R., Seegmiller, A. C., Douds, J., Compton, M., Szentirmai, E., Woodruff, M., Huang, W., Stratton, C., & Eichbaum, Q. (2018). The Impact of Speed and Bias on the Cognitive Processes of Experts and Novices in Medical Image Decision-Making. Cognitive Research: Principles and Implications, 3, 28. https://doi.org/10.1186/s41235-018-0119-2 Humr, S., Canan, M., & Demir, M. (2025). A Quantum Probability Approach to Improving Human–AI Decision Making. Entropy, 27(2), 152. https://doi.org/10.3390/e27020152 Kvam, P. D., Busemeyer, J. R., & Pleskac, T. J. (2021). Temporal oscillations in preference strength provide evidence for an open system model of constructed preference. Scientific Reports, 11(1), 8169. https://doi.org/10.1038/s41598-021-87659-0 ================================= Secondary Sources: APA on AI companion app growth and youth risk survey Psychiatric Times on AI romantic relationships MIT Media Lab profile of AI relationship users Nolo on 2026 chatbot suicide/violence lawsuits Bloomberg on ~40 tracked chatbot harm lawsuits TorHoerman Law on FSU shooting survivor's OpenAI lawsuit ConsumerNotice.org on chatbot self-harm lawsuits Sokolove Law on Character.AI/Google January 2026 settlements Callfob 2026 chatbot lawsuit legal overview Malay Mail on Adam Raine's death and OpenAI lawsuit CBS News on Zane Shamblin ChatGPT lawsuit Fox News on Lauren Bannon's ChatGPT cancer diagnosis