Evidano is an AI-powered qualitative data analysis platform that helps research teams ingest, align, and analyze interview transcripts, device logs, and surveys. Detecting distress before a person reaches out is the problem UbiMyTherapist aims to solve. A University of Ottawa prototype, reported June 28, 2026, combines heart-rate variability, speech tone, and text to build a digital twin and offer proactive support. For researchers and UX teams, the payoff is a reproducible approach to fusing biosignals and qualitative context so teams can surface early warning signals and design timely interventions. This post translates that study into a practical workflow for qualitative analysis of wearable data and shows how to operationalize it in Evidano, secure, research-first tooling that combines thematic, cross-segment, and frequency analyses with AI chat over your corpus.
Key Takeaways
UbiMyTherapist is a University of Ottawa research prototype that fuses heart-rate variability, speech-tone features, and text to detect distress and support proactive interventions; its reactive mode was validated with 24 participants and scored higher on empathy and personalization versus standard LLMs.
- Published June 28, 2026, the prototype used heart-rate variability (HRV), speech-tone changes, and written text as signals.
- The reactive mode was evaluated with n=24 participants, and licensed therapists rated its responses as more empathetic and personalized than baseline LLM setups.
- UbiMyTherapist is a research prototype, not a consumer app, and proactive real-time wristwatch interventions remain under development.
- Researchers can reproduce similar pilots by aligning timestamps, fusing biosignals with transcripts, and validating outputs with clinicians.
Fast take + source
Fast take: University of Ottawa researchers prototyped UbiMyTherapist, an AI assistant that reads smartwatches, earbuds, and phone text to detect distress and intervene proactively, and the study was reported in The Next Web.
- Published: June 28, 2026; sample: n=24 (reactive evaluation).
- Signals used: heart-rate variability, speech-tone changes, written text.
- System: a digital twin combining medical and psych history with live signals.
- Status: research prototype, not a consumer app.
Findings snapshot
| Metric / Item | Value | Source / Note |
|---|---|---|
| Published | June 28, 2026 | University of Ottawa prototype reported in The Next Web |
| Participants (reactive eval) | 24 | Licensed therapists assessed therapeutic soundness |
| Input signals | HRV, speech tone, text | Used to infer emotional state |
| Modes | Reactive + Proactive | Reactive tested; proactive under development |
| Outcome reported | Higher empathy & personalization vs. standard LLMs | Therapist ratings (study) |
| Product status | Research prototype | Not a consumer app |
What happened, method in plain English
What happened: the research team built a pipeline that ingests biosignals (like heart-rate variability from a watch), paralinguistic features from earbuds (changes in speech tone), and text inputs, then fuses those streams into a running profile or digital twin that adds clinical and psych history to live emotional signals.
- The assistant runs two modes: reactive, which responds when users ask, and proactive, which monitors for distress and intervenes; the reactive mode was validated with 24 participants.
- Signal fusion matters: combining physiological and textual cues increases context for responses and helps shape personalized interventions.
- Human-in-the-loop validation matters: clinicians evaluated therapeutic appropriateness and found higher empathy and personalization versus standard LLM-based chatbots.
- Prototype caveat: proactive, real-time wristwatch interventions are still being developed and require safety and regulatory validation before deployment.
Implications for researchers, UX teams, and analysts
For qualitative researchers
For qualitative researchers: the primary opportunity is to link subjective reports, like diaries and interview transcripts, to passive biosignals and code for pre-crisis patterns.
The UbiMyTherapist study shows small-n prototyping (n=24) can validate whether fused signals improve interpretability before scaling.
Research tip: capture timestamps and device metadata so you can align transcripts with physiological spikes during coding.
For UX/product teams
For UX and product teams: proactive interventions change consent flows, notification design, and escalation paths and therefore require staged prototyping.
Design implication: use reactive modes first and therapist-reviewed scripts before enabling push interventions to manage safety and user trust.
Measure both experience and safety: empathy ratings from clinicians are a useful early metric, but real-world pilots need behavioral and retention KPIs.
For clinical & policy analysts
For clinical and policy analysts: the system should be positioned as augmentative, not a replacement for therapy, and proof of clinical utility is required before deployment.
Regulatory note: passive biosignal inference faces validity questions and needs clinical validation and clear escalation paths to human care.
Ethics in one line: design for consent, opt-outs, and clear escalation to human clinicians for safety.
FAQ: qualitative analysis of wearable data
What signals did UbiMyTherapist use to detect distress?
UbiMyTherapist used heart-rate variability, speech-tone changes, and written text as its primary input signals.
The study fused physiological data from smartwatches, paralinguistic features from earbuds, and phone text to infer emotional state and build a digital twin.
How was UbiMyTherapist evaluated?
The reactive mode was evaluated with 24 participants and licensed therapists compared its responses to standard LLM-based chatbots.
Therapists rated the reactive mode higher for empathy and personalization according to the report.
Is UbiMyTherapist available as a consumer app?
No, UbiMyTherapist is a research prototype and not a consumer application.
The authors report that proactive, real-time wristwatch interventions are still under development and require further validation.
How can researchers run a wearable-informed qualitative pilot?
Researchers can run a pilot by aligning timestamps across device logs, audio, and transcripts, and following a reproducible analysis pipeline with clinician validation.
The post includes a seven-step checklist that starts with defining outcomes and safety pathways and finishes with piloting reactive messaging only until safety is validated.
Do more, faster with Evidano (map to this use case)
Ingest and align multimodal inputs
Ingest and align multimodal inputs by importing transcripts, timestamped device logs, and survey sheets and aligning them by timestamp for unified coding and analysis.
Evidano lets teams import interview transcripts, timestamped device logs, and survey sheets and align artifacts for unified coding.
Thematic and cross-segment analysis
Run thematic and cross-segment analysis to spot patterns across small cohorts and segments like age and device type.
Evidano runs thematic and frequency analyses and cross-segment comparisons to surface which signals co-occur with coded distress themes.
Human-in-the-loop validation
Ensure therapeutic appropriateness by validating outputs with clinicians and capturing annotated codebooks for reviewer sign-off.
Evidano exports clinician-annotated codebooks, supports AI-assisted re-coding, and produces side-by-side examples for reviewer sign-off.
Secure research & follow-ups
Protect sensitive health data and control model training by using secure, research-first tooling with explicit guarantees about data use.
Evidano provides end-to-end encryption and a guarantee that customer data is never used to train third-party models and supports AI avatar interviewers for safe follow-ups.
Explainability & visuals
Translate signal fusion into stakeholder-ready reports with explainable visuals and interactive queries over the corpus.
Evidano offers one-click visualizations, such as word clouds, co-occurrence networks, and hierarchical code trees, plus an AI chat over your corpus for ad-hoc queries.
Checklist: run a wearable-informed qualitative pilot (7 steps)
Checklist: run a wearable-informed qualitative pilot in seven steps.
Step 1: Define outcome and safety pathways (what counts as a distress event and escalation rules).
Step 2: Collect synchronized data, ensure timestamps on device logs, audio, and text entries.
Step 3: Import artifacts into Evidano, including transcripts, device CSVs, and surveys.
Step 4: Create a minimal codebook with clinician input and run AI-assisted initial coding.
Step 5: Run cross-segment frequency and co-occurrence analyses to surface signal-theme links.
Step 6: Validate outputs with licensed clinicians and iterate prompts and the codebook.
Step 7: Pilot reactive messaging only, and graduate to proactive monitoring only after safety validation and IRB or regulatory checks.
Wrapping up: what to do next
Wrapping up: UbiMyTherapist illustrates that fusing biosignals and qualitative context can make support more timely and personalized, but this approach requires careful design, clinician validation, and reproducible analysis pipelines.
If you are running pilots or evaluating passive-device signals, start by aligning timestamps and running thematic and cross-segment analyses on a small cohort using tools built for qualitative rigor.
Ready to try it? Try Evidano for free.