The Prosthetic Mind: A Case for Always-On AI Cognitive Companions for Profound Autism — and an Invitation to a Necessary Controversy

Roughly one quarter of the autism spectrum consists of individuals with profound autism: minimally verbal or non-verbal people, frequently with co-occurring intellectual disability, who will require intensive support for the entirety of their lives. For this population, the dominant paradigms of assistive technology — skill-training apps, therapy games, social-skills tutors — rest on an assumption that is, for them, largely false: that the technology’s job is to teach, and then get out of the way. This paper argues for something the field has been reluctant to say aloud: for people whose core capacity for real-time social cognition will never develop, we should build machines that perform that cognition for them, permanently, in real time. I propose the concept of an always-on, body-worn cognitive prosthesis: a multimodal AI system that continuously perceives the wearer’s physical and social environment, reasons about it using a large language model, and delivers simple, moment-to-moment behavioral guidance — say hello now; wave; step back; this person is leaving, say goodbye — through a constrained, clinician-approved action vocabulary. I defend this proposal through the ethics of prosthetics: we do not demand that a wheelchair teach its user to walk, and we should stop demanding that cognitive assistive technology justify itself through remediation. I then state, at full strength, the four objections that should alarm any reader — the annexation of a human being’s agency by a machine, the conversion of care into surveillance, the automation of neurotypical masking, and the dignity of risk — and I concede that not all of them can currently be answered. The purpose of this paper is not to close this debate but to force it into the open. If such systems are technically feasible within this decade — and I argue they are — then the community’s silence is not caution. It is abdication.