Collecting language, speech acoustics, and facial expression to predict psychosis and other clinical outcomes: strategies from the AMP® SCZ initiative
Bibliographic record
Abstract
Speech-based detection of early psychosis is progressing at a rapid pace. Within this evolving field, the Accelerating Medicines Partnership® in Schizophrenia (AMP® SCZ) is uniquely positioned to deepen our understanding of how language and related behaviors reflect early psychosis. We begin with detailed standard operating procedures (SOPs) that govern every stage of collection. These SOPs specify how to elicit speech, capture facial expressions, and record acoustics in synchronized audio-video files-both on-site and through remote platforms. We then explain how we chose our sampling tasks, hardware, and software, and how we built streamlined pipelines for data acquisition, aggregation, and processing. Robust quality-assurance and quality-control (QA/QC) routines, along with standardized interviewer training and certification, ensure data integrity across sites. Using natural language processing parsers, large language models, and machine-learning classifiers, we analyzed Data Release 3.0 to uncover systematic grammatical markers of psychosis risk. Speakers at clinical high risk (CHR) produced more referential language but fewer adjectives, adverbs, and nouns than community controls (CC), a pattern that replicated across sampling tasks. Some effects were task-specific: CHR participants showed elevated use of complex syntactic embeddings in two elicitation conditions but not the third, underscoring the importance of the language sampling task. Together, these results demonstrate how computational linguistics can turn everyday speech into a scalable, objective biomarker, paving the way for earlier and more precise detection of psychosis.Video Link: https://vimeo.com/1112291965?fl=pl&fe=sh.
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 distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.001 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
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 teacher head, 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".