A Master thesis topic is available at the Centre for Digital Health Interventions!

What this thesis is about

Valico’s care companion is built as a modular adaptive engine, a just-in-time adaptive intervention (JITAI, after Nahum-Shani 2018): it senses the dyad’s state and delivers the right support at the right moment. Rather than test the whole engine, this thesis takes one module and develops it end to end, from design specification through a working prototype to a first evaluation.

The chosen module is the caregiver-strain support lever: sense when a caregiver’s strain is rising and when they are open to engage, then deliver a brief, evidence-based coping interaction (CBT / ACT) in voice or chat. It is the best-evidenced part of the engine (Hounsri 2024, 19 RCTs), which lets the thesis concentrate on the harder, more original questions: how to sense the right moment and how to design the interaction, rather than re-proving that the content works.

What we’re aiming for

Produce a fully specified, working, and evaluated single JITAI module. Concretely:

• Translate the conceptual model’s caregiver-strain lever into an implementable specification: decision points, tailoring variables for vulnerability (rising strain) and receptivity (a good moment to reach out), graduated intervention tiers, and if-then decision rules, with an always-on safety and escalation path.
• Build it as a working voice-and-chat module on Valico’s agent platform.
• Evaluate its feasibility and acceptability with caregivers, and test whether its strain signal (a one-item check-in plus vocal-stress from spoken turns) tracks a validated burden measure.

What you will do

• Specify the module as a JITAI: decision points, vulnerability and receptivity tailoring variables, graduated intervention tiers, decision rules, and a safety / escalation path.
• Build a working prototype of the module on Valico’s existing agent platform (design and configure, not build an agent from scratch).
• Define and instrument the measures: signal quality (does the strain signal track ZBI-12 / PHQ-2), acceptability (Theoretical Framework of Acceptability), usability (System Usability Scale), and correct, safe escalation.
• Run a small evaluation with caregivers (think-aloud and acceptability sessions), recruited via FRAGILE Suisse or a clinical site.
• Analyse the results and recommend the next design iteration and a micro-randomized trial (MRT) plan.

Who will participate

You will work in collaboration with Valico AG, a Swiss digital health startup developing advanced AI tools for stroke care. You will design and build directly on its platform, giving rare hands-on experience turning a clinical conceptual model into a working digital intervention.

What you will gain

• Learn to design a just-in-time adaptive intervention end to end, from theory to a working artifact.
• Hands-on conversational-AI and voice-agent development in health.
• Mixed-methods evaluation experience (signal quality plus acceptability)
• A thesis with both a built artifact and a study behind it
• Insight into the JITAI model and into real-world startup-academic collaboration.

What we’re looking for

A motivated Master’s student with an interest in digital health, human-computer interaction, or conversational AI, and some comfort with prototyping or configuring software. Enrolment in a relevant Master’s programme (health technology, human-computer interaction, computer science, biomedical engineering, health sciences, or a related field) is required.

Start: As soon as possible / by arrangement

Duration: 6 months (or to be discussed)

Supervision: Prof. Dr. Marcia Nißen (CDHI), Dr. med. Eugenio Abela (Valico AG, clinical lead)