Follow-up after digital consultations — Technical https://thesis.uya.no/proposals/close-the-follow-up-gap/#technical-track BACHELOR Final result: A working follow-up checklist with tested validation rules. Task: Build a structured follow-up form that links actions to their source note. Data: 30 fictional consultation notes (planned) Write 150–250 words per note with 0–4 actions. Label action, owner and deadline separately; include 10 notes with missing owners or dates. Requires: Clinical review of scripts; recruit students. No patient notes. Method: Enter actions manually from the fictional notes; check missing-owner/date handling and observe 4 task sessions. STEPS Build and test 1. Read the starting sources and choose one established implementation method. 2. Write the requirements, data fields, system diagram and test cases. 3. Prepare 30 fictional consultation notes. Make the answer key and pass criteria before testing. 4. Build a working version of a working follow-up checklist with tested validation rules. 5. Run function, integration and failure-case tests. Record each result. 6. Run the practical evaluation and list the changes the system still needs. MASTER Final result: Pinned extraction pipeline, confirmation interface, labelled notes and four-way evaluation. Research question: Study whether confirmation reduces extraction errors without creating false responsibility. Compare keyword/date rules, raw Qwen, manual extraction and Qwen plus confirmation; analyse missing owners and dates separately. Research result: Evidence about how confirmation changes accuracy and accountability in follow-up work. STEPS Design, build and test (DSR) 1. Read the newest papers and list the closest existing systems. 2. Write down the versions, fields, data split, case assignment, random seeds and correct answers for 30 fictional consultation notes. 3. Draw the user workflow, data model and system architecture. List the requirements and pass criteria. 4. Build a working version of pinned extraction pipeline, confirmation interface, labelled notes and four-way evaluation. 5. Test every function, connection and failure case. Save the failed tests as well as the passed tests. 6. Compare the system with the named alternative. Then run the user task or decision task in the assignment. 7. Report the measured result, the failed cases and the design lessons another team can reuse. CURRENT PROJECT LITERATURE Evaluating Large Language Models in extracting cognitive exam dates and scores (2024, peer-reviewed): https://doi.org/10.1371/journal.pdig.0000685 Provides a recent expert-reviewed clinical information-extraction design and error analysis. Systematic Evaluation of Large Language Models for Post-Discharge Clinical Action Extraction (2026, preprint): https://arxiv.org/abs/2605.06191 Directly studies the emerging task of extracting safety-critical follow-up actions from notes. IS THEORY STARTING POINTS Amershi et al. (2019) — Guidelines for Human-AI Interaction: https://doi.org/10.1145/3290605.3300233 Select and justify the correction, explanation or uncertainty design being tested. Buçinca et al. (2021) — To Trust or to Think: https://doi.org/10.1145/3449287 Derive a competing explanation based on verification effort and overreliance. Search Scopus or Web of Science and ACM Digital Library using the topic query, then follow citations to the thesis start date. Record searches and compare methods, data, findings and limitations in literature-matrix.csv. Use that review to confirm or revise the gap and choose a current comparator. The linked papers are starting points. PILOT QUESTION Does source-linked task confirmation reduce missed or incorrectly assigned follow-up actions? EXAMPLE TOOLS Qwen2.5-7B-Instruct, Python, Streamlit Equivalent tools are fine. FIRST THREE ACTIONS 1. Write 3 fictional notes and manually label every action, including explicit missing values. Confirm access to inference compute and record model, software and hardware versions. 2. Create the development/held-out split before trying prompts. Keep held-out results hidden until tuning is complete. 3. Pilot action confirmation on 3 development notes and fix the scoring rules. LITERATURE SEARCH clinical consultation follow up task extraction accountability DATA 30 fictional consultation notes (planned) Write 150–250 words per note with 0–4 actions. Label action, owner and deadline separately; include 10 notes with missing owners or dates. Qwen2.5-7B-Instruct: https://huggingface.co/Qwen/Qwen2.5-7B-Instruct Transformers generation settings: https://huggingface.co/docs/transformers/main_classes/text_generation HOW TO TEST IT On 20 held-out notes compare keyword/date rules, raw Qwen, manual structured extraction and Qwen plus human confirmation. Report action-span precision/recall/F1, owner/date accuracy conditional on matched actions, invented responsibility, residual post-confirmation errors, corrections introduced, source use and time. FINAL RESULT Pinned extraction pipeline, confirmation interface, labelled notes and four-way evaluation. ACCESS OR PEOPLE Clinical review of scripts; recruit students. No patient notes. TEMPLATES https://thesis.uya.no/starters/close-the-follow-up-gap/technical/data-plan.csv https://thesis.uya.no/starters/close-the-follow-up-gap/technical/evaluation.csv https://thesis.uya.no/starters/close-the-follow-up-gap/technical/literature-matrix.csv Use participant codes instead of names or email addresses.