O-467 Evaluating the impact of sex and gender on the performance of machine learning for auto encoding of job titles
Bibliographic record
Abstract
Introduction Ongoing studies into the use of algorithms for the automated coding of job titles to the Canadian National Occupation Classification have performance accuracy which are at least equivalent to manual coding accuracy. Moreover automated coding provides significant time savings. These studies have identified that both natural language processing and machine learning algorithms are effective for auto coding. Whereas NLP based and machine learning approaches both rely on bespoke rules, and existing data sets, machine learning models can proliferate bias from training data if not corrected. Objectives The goal of the study is to explore the impact of altering sex/gender ratios in training data sets on overall performance of the machine learning based prediction of NOC codes using patient provided job titles. Methods Using data participant patient data provided by Atlantic PATH, training data sets were prepared for 100 4-digit NOC categories. The data sets were prepared with sex/gender ratios of 50/50 30/70, 70/30. The data sets were used to train ENENOC machine learning platform and tested on a set of manually coded job titles provided by Atlantic PATH CanPATH . Performance levels were contrasted for all 4-digit NOC categories used in the study. Results Initial results in this preliminary study have identified that sex and gender are variables that can influence auto coding performance, however the extent to which overall coding accuracy is impacted is relative minor. Further studies are required with larger training sets to fully explore the extent of sex and gender as contributing variables to bias to ENENOC. Conclusion We initiated studies to investigate the impact of sex and gender bias on performance of the ENENOC algorithm. Together, the ENENOC contributed training and test sets provide a suitable framework for ongoing work in this area.
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 | 0.005 | 0.027 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.000 | 0.001 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.006 | 0.002 |
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".