01 · Respond
Affective responsiveness
What should AI say?
Design communication that recognises emotion, interprets the person's situation and explores their needs appropriately.
Output: Empathic Response Assistant →Translating trustworthy AI into equitable and safe mental health care for vulnerable populations.
Empathic language, safe emotional regulation and trustworthy implementation are related, but they are not the same achievement. TASH tests them in sequence.
Design communication that recognises emotion, interprets the person's situation and explores their needs appropriately.
Output: Empathic Response Assistant →Define when an emotional response should persist, recover, escalate or be constrained across changing evidence and crisis states.
Output: Affective Safety Controller →Study appropriate reliance, mind perception, equity, accountability and implementation in real care pathways.
Output: Implementation framework ↓The first layer turns affective explainability into an observable interaction skill. The RP examines why emotionally responsive explanations may shape trust through perceived Experience; the prototype makes the communication itself testable.
One empathic reply does not guarantee a safe conversation. The second layer asks whether AI can carry emotional context forward, respond quickly to danger and recover only when credible safety evidence appears.
The final layer returns to the healthcare management question: when does affective AI deserve reliance, for whom, and under which institutional safeguards?
Cognitive explanations can reduce decisional conflict without closing the emotional trust gap.
Test whether affective explainability changes perceived Experience rather than only Agency.
Measure calibrated trust, willingness to comply and observable decisions without encouraging over-trust.
Examine accessibility, subgroup differences, unequal failure costs, oversight and accountability.
Operationalise empathic communication and establish LLM baselines.
Build SNN/Brian2 temporal dynamics and test safety-relevant behavioral constraints.
Evaluate mind perception, adoption, compliance, equity and organisational deployment.