FER and Peer Evaluations for 27 Pre-Service Teachers in Microteaching

Shota Shirasaka's study from 2026 provides data on the disconnect between automated facial expression recognition and human peer evaluations. It includes 702 peer ratings across nine dimensions for 27 Japanese pre-service teachers conducting 3-minute microteaching sessions. The dataset, shared under a CC-BY-4.0 license, contains results from DeepFace and Py-Feat analyses organized into three evidence levels.

Use Cases

• Testing correlations between automated FER outputs and human ratings based on the nine evaluation dimensions.
• Analyzing halo effects in peer evaluation based on the described principal component analysis showing strong structure.
• Investigating the impact of forced distribution rating designs on statistical conclusions.
• Comparing results from different FER tools (DeepFace vs. Py-Feat) on the same microteaching video data.

Strengths

• Includes 702 peer evaluation ratings, providing a substantial number of human judgments.
• Applies a structured, multi-level analytical framework (Level 1-3) to test the null hypothesis.
• Dataset is openly licensed under CC-BY-4.0, facilitating reuse and verification.

Limitations

• Row and column counts are unknown, making it difficult to assess the dataset's structure and scale prior to download.
• The dataset is very small at 127.8 KB, indicating limited scope and likely summary statistics rather than raw observation-level data.
• Column-level documentation is absent; field semantics must be inferred from the accompanying research paper.

Provenance

Source

Shota Shirasaka via figshare.

Collection Method

Data collected from 27 pre-service teachers who conducted microteaching sessions, with peer evaluations and subsequent automated facial expression analysis.

Freshness

Last updated 2026-06-03 06:01:15.

Geography

Study conducted with Japanese pre-service teachers; specific geographic scope within Japan is not detailed.