# Reviewer Objection Bank

## Trace data

- Why should this platform event indicate the learning construct?
- Which learner actions were not captured by the system?
- Did the analysis preserve timing, sequence, and task context?

## Text and discourse

- Who segmented the text and how stable were the segments?
- Were automated codes validated against human judgment?
- Are frequency counts being overinterpreted as quality?

## Social networks

- What exactly counts as a tie?
- Are missing actors or off-platform interactions distorting the network?
- Does centrality have a learning-theoretic interpretation here?

## Video and multimodal data

- Why were these episodes selected?
- How were streams synchronized?
- What are the camera, microphone, and sensor blind spots?

## Prediction

- Was the model evaluated on held-out data?
- Could outcome information leak into predictors?
- Are subgroup performance and fairness reported?

## Intervention claims

- What assignment or comparison supports the effect claim?
- Were attrition, fidelity, and contamination checked?
- Is the wording causal, associative, or descriptive?