SYNTHETIC TRAINING DATA
Purpose
Practise moving from an extraction matrix to codes, themes, a synthesised results paragraph and evidence-based research gaps — using a small, shared set of six fictional studies. Everyone works on the same data, so the group can compare themes and discuss differences in interpretation.
Practice review question
RQ (practice): How does AI-generated feedback influence novice programmers’ performance and help-seeking behaviour, and what limits the current evidence?
Files
- Synthesis practice matrix (CSV) — pre-filled extraction for six synthetic studies SYN-01 to SYN-06.
- Full synthetic texts for SYN-01, SYN-02 and SYN-03 (from Stage 2).
- Theme coding matrix (blank)
Activity (45–60 minutes)
- Individually (15 min): read the “Key findings” and “Limitations” columns. Write 8–12 short codes with study IDs.
- In pairs (15 min): merge your codes and group them into 3–4 descriptive themes. Write a one-sentence definition for each theme.
- In pairs (10 min): for each theme, list supporting and contradicting studies and judge the strength of evidence using the “Quality” column.
- Individually (10 min): write one synthesised results paragraph for your strongest theme and two evidence-based research gaps.
- Whole group (10 min): compare themes. Where did pairs disagree, and why?
Your practice notes
Saved locally as you type
Debrief questions
- Did your paragraph compare studies, or describe them one by one?
- Which claims would change if you excluded the lowest-quality studies?
- Is “students value AI feedback” a finding about learning, or about perception? How should it be worded?
Facilitators: a suggested theme structure is in practice-pack/facilitator-theme-key.csv. See the facilitator notes.