This repository contains the code, survey data, and analysis scripts accompanying the working paper "Data Annotations as Pedagogical Hints: From Subjective Labels to Critical Thinking".
When machine learning courses use pre-labeled datasets, the subjectivity inherent to human annotations is hidden from students' perception. This paper explores how tasking students with manual data annotation teaches them how human judgement shapes AI training data.
We implemented an annotation module at two universities β Fontys University of Applied Sciences (NL) and IT University Copenhagen (DK) β where students annotated skin lesion images for hair coverage on a 3-point scale. Survey data from 43 students was collected and analysed.
- To what extent does annotating data in groups affect students' perception of data quality and its effect on bias and fairness in AI models?
- From the perspective of students and educators, what are the perceived pedagogical strengths and limitations of using data annotations as pedagogical?
| Folder / File | Description |
|---|---|
Data_Annotations_as_Pedagogical_Hints__DAPH_.pdf |
Working paper |
0_data/ |
Survey data (soscisurvey), hair annotions (ITU), and coded responses (doccano) |
1_code/ |
Quantitative analysis scripts (Likert distributions, Fleiss' ΞΊ) and qualitative thematic analysis |
If you use this code or data in your research, please cite our paper: [will follow]