Generation and Classification of Motivational-Interviewing-Style Reflections for Smoking Behaviour Change Using Few-Shot Learning with Transformers
Bibliographic record
Abstract
If conversational agents can take on a therapeutic role, they may provide a scalable way to help many people suffering from addictions. Motivational Interviewing (MI) is a validated therapy for behaviour change that can be applied to addiction, including smoking cessation. A core technique in MI (and many other kinds of talk therapy) is to pose an open-ended question concerning a negative behaviour, and then to provide a reflection of the response. Reflections can be a simple restatement of the response, or a more complex inference from prior statements or general knowledge, and they help someone contemplate the behaviour more deeply. We describe a method to generate reflections that uses few-shot priming of the GPT-2 and GPT-3 language models. These produce very promising simple and complex reflections, but also some that are off-topic or irrelevant. To filter these, we train a classifier to detect poor reflections, employing samples labeled by an MI expert. Its accuracy is 81%, sensitivity 90% and specificity 71%. We show that GPT-2 can generate acceptable reflections at a 54% success rate, and when combined with the classifier/filter produces acceptable reflections 73% of the time. The GPT-3 model has a native success rate of 89%.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.013 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.003 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".