Six conversations that redirected the project

Core findings — from his own report, safe to use:

• Music demonstrably helps dementia patients, but the mechanism is memory-linked. The benefit comes from familiar music of the patient's youth — songs tied to autobiographical memory. Purpose-composed "treatment music" or generic relaxing audio shows little effect in his experience. → An AI-generated-new-music product has weak fit for dementia; a dementia product would need era/biography-personalized licensed catalogues.
• Monetization in China is difficult for this population. The patient is not the buyer; families prioritize medical treatment and physical care spending; elderly users have low digital adoption and near-zero willingness to pay for apps. He could not judge Europe/US payer systems. → the dementia option scores low on viability in the China market.
• He does not think in commercial terms — his lens is clinical benefit and care burden, not product/market. (Honest limitation: the business assessment is the researcher's, not the expert's.)

Additional themes — [VERIFY] each before use, not yet confirmed by the interviewee:

• [VERIFY] Music's clearest observable effects are on agitation and mood (e.g. calming during evening restlessness) and on engagement (patients who barely speak may sing along to old songs) — more than on cognition itself.
• [VERIFY] Effects are transient — benefits appear during/around listening sessions and fade; music is care support, not disease modification.
• [VERIFY] Caregivers, not patients, would be the practical user of any digital tool; caregiver burden is the real pain point.
• [VERIFY] Personalization is labour-intensive today (families/nurses guessing what the patient loved decades ago); automating "find the songs of their youth" would be the useful part.
• [VERIFY] On anxiety more broadly: he sees anxiety symptoms co-occurring in patients and especially in family caregivers, and regards music/relaxation as a reasonable low-risk self-care adjunct — while stressing that moderate-to-severe anxiety disorders need professional treatment and a consumer product should not claim therapy.
• [VERIFY] Hospitals/institutions in China are unlikely buyers (procurement driven by medical devices and staffing, not wellness software).

Researcher's interpretation: kills the "AI-generated music for dementia" route on two criteria of the convergence matrix — evidence-fit (familiar-music mechanism ≠ our AI-new-music asset) and viability (payer problem in China). Redirects attention to adults with everyday anxiety/stress: self-aware, self-paying, digitally native, measurable.

Key findings:

• Medication adherence is the daily battle; she hides crushed pills in food and lies to get compliance. Caregiver burden is the real pain, not "content". [D]
• Tried community-recommended music (old patriotic songs, 邓丽君). Verdict: calms her mother for ~10 minutes then the effect vanishes; she sees it as 心理安慰/噱头 (placebo/gimmick) — "like a toy that distracts a child — not treatment". [D]

Implication: confirms Dr. Wang — familiar music = brief distraction, not therapy; reinforces low fit of an AI-music product for dementia. [V]

Key findings:

• Sundowning (evening agitation/delirium) is the worst window; medication given via disguise. [D]
• Bought a player for 越剧 (opera). Effect is unpredictable: sometimes she falls asleep, sometimes percussive passages over-arouse her — shouting, neighbours complain. Stopped using it; relies on sedatives during real episodes. [D]

Implication: music can worsen agitation if not the right stimulus — "unpredictable" is a recurring word. Personalization + arousal-matching matters more than "music" per se.

Key findings:

• Anxiety = "always on call" dread; notification sounds trigger a somatic response (heart skips, chest tightens). [D]
• Wears noise-cancelling headphones 10+ hrs/day. Can't tolerate lyrical pop when anxious; late-night algorithmic "emo/sad" recommendations deepen the anxiety ("existential crisis, want to quit"). Metal/EDM → more agitated. Nature white noise has jarring sharp sounds (bird calls, high piano). [D]

Key findings:

• "Forced wakefulness": body exhausted, cortex "setting off fireworks", ruminating on the day; racing heart; awake till 3–4am. [D]
• Standard "助眠歌单" (sleep playlists) ineffective or counterproductive. The sweet "治愈系/healing" genre triggers reactance — feels 虚伪 (fake). Vocal humming makes her more awake; strong rhythm → her heart entrains to the beat (wrong direction). [D]

Key findings:

• Cash-flow/payroll stress; chronic "low-oxygen" chest tightness, irritability; anxiety "floods in like a tide" alone at night. [D]
• Paradox: cannot tolerate absolute silence (mind races) — must have sound. But highly sound-sensitive: sad strings → "company's going bankrupt tomorrow"; happy pop → feels worse/annoying; brainwave (Alpha) audio with a low 嗡嗡 hum → physical nausea. [D]
• Current substitute: air purifier on max.