Sweet Nothings: The Procrustean Therapist Pretends to Listen

“Science can overcome all things, even human emotions.”

A venerable shrink joke features a freshly anointed analyst, shaken by the volcanic outflow of nasty revelations and primary process bile during his therapeutic hours, who envies a sleek older colleague down the hall who always emerges from his consulting room seemingly unruffled. At long last, no longer able to contain himself, the young anxious analyst implores the dapper savant to divulge the secret to serene composure.

“I don't listen.”

Neither does AI but, as in the case of the oblivious analyst, the clientele may not always catch on. When MIT computer scientist Joseph Weizenbaum in the 1960s puckishly rolled out ELIZA, a Rogerian-style therapeutic programme with its kneejerk knack for reframing 'patient” statements as if they were sensitive probing questions, he was gobsmacked that some colleagues became engaged with, and even attributed human agency to, the poseur seer.1 This alarmingly widespread inclination to endow pseudo-articulate machines with souls became dubbed the ELIZA effect, and AI hype is its present-day culmination.

Large Language Model gadgets scoop up, cross-tabulate and algorithmically rearrange Niagras of data but they remain incapable of listening, as a host of critics note, in anything ‘other than a figurative sense of the verb.‘2 In her musings Piotrowska raises, or resurrects, a diabolical issue attending every gizmo ever designed to parallel, augment or surpass human capability, and inevitably in doing so, brush aside indigestible and therefore supposedly irrelevant human elements.3 Whatever does not fit does not count. Success, in however parched a form, depends on imposing reductive formats on human consciousness, which never goes well. No matter how many expensive (usually for other than the hype merchants) disappointments pile up, the allure of technical fixes never fades entirely both because it stirs childhood fan