When a support system breaks, its real incentives show. A field report on affective blindness, governance, and why bots escalate hostility but never harm.
There is a moment in every system where its true nature reveals itself—not in its marketing copy, not in its onboarding flow, but in the instant something breaks and a human being reaches for help.
That moment arrived for me this week on Handshake, a platform explicitly designed for students and alumni. A fraudulent employer slipped through their verification pipeline, harvested personal information, and triggered the kind of identity-theft cascade that no student should ever have to navigate. When I attempted to report it, the "Submit Feedback" button simply didn't work. Not in Firefox. Not in Edge. Not in private mode. Not at all.
The support chatbot that followed was fast, polished, and utterly blind. It ignored every emotional indicator I threw at it—urgency, distress, frustration, escalation—and looped me through irrelevant flows until I finally told it, bluntly, that it was useless. Only then did it escalate to a human.
And in that moment, the system revealed its architecture.
