Conversational Breakdown Detector for a Motivational Interviewing Conversational Agent
Bibliographic record
Abstract
A conversational breakdown in human-chatbot interaction refers to a disruption or failure in the communicative flow between the human user and the chatbot. To recover a disrupted conversation, the first step is to detect the breakdown. Researchers have proposed methods using supervised learning and semi-supervised learning in dialogue systems to achieve the goal of detecting conversational breakdown. However, few studies have focused on detecting breakdowns in automated therapeutic conversations, especially conversations led by motivational interviewing chatbots. The presence of conversational breakdowns has negative impacts on the human-chatbot interaction, such as frustration, dissatisfaction, or loss of trust. This gap suggests a need to build a robust and efficient conversational breakdown detector that recognizes interruptions during the conversation. Conversational breakdown detection paves the way for further action to recover conversations. In this paper, I develop a novel, unifying framework called “CIMIC” for characterizing the conversational breakdowns of “MIBot,” a motivational interviewing conversational agent for smoking cessation. I collect 200 pieces of conversational data through Prolific and annotate them using the CIMIC framework with a group of four trained annotators. The annotated dataset is then applied as the training set to fine-tune GPT-3 models to build a conversational breakdown detector for the MIBot.
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 | 0.005 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.000 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.002 | 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".