Development of an Artificial Intelligence System to Monitor Digital Marketing of Unhealthy Food to Children: Research Protocol
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Unhealthy food marketing to children adversely affects their diet quality and health. The negative impacts of this marketing may be amplified on digital media, which allows industry to use artificial intelligence (AI) to market unhealthy food to children in covert ways. Health Canada is developing regulations to prohibit digital marketing of unhealthy food that appeals to children <13 years. However, reliance on adults to manually assess food marketing to children on digital media has limited understanding of key targets for policy and capacity to monitor policy adherence. To address these gaps, we are developing an AI system to monitor marketing of unhealthy food to children on digital media, including websites, YouTube, social media and mobile gaming apps. Our web and mobile scrapers continuously collect marketing instances that may be viewed by individuals in Canada on websites and social media applications popular with children. This has allowed us to accumulate a database of > 615,000 marketing instances. The AI system extracts features from each marketing instance to determine whether foods are present, and if so, whether they are unhealthy according to Health Canada's standards (based on the presence of added saturated fat, added sodium and/or free sugars). Next, the AI system uses a supervised machine learning model to assess whether child appealing marketing techniques are present. In the final step, the system integrates all of the data collected to determine whether a given marketing instance features unhealthy foods and appeals to children. The system can be applied to monitor the extent and nature of digital food marketing to children internationally. It can also be retrained to monitor adherence to country-specific policy. This is a protocol paper so there are no results. The AI system provides a scalable, objective and reproducible manner to identify digital marketing of unhealthy food that appeals to children across the digital marketing landscape. The system can assist researchers and policy makers to study children's exposure to digital marketing of unhealthy food and its impacts, and to monitor adherence to policy that restricts this marketing. Canadian Institutes of Health Research.
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.
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 | 0.002 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.001 | 0.001 |
| 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 it