# Claim Wording Clinic

Use this sheet before writing findings, abstracts, slide titles, or paper claims.

| Risky sentence | Why it overclaims | Safer replacement |
|---|---|---|
| The dashboard improved learning. | Needs causal design and outcome evidence. | Learners using the dashboard showed higher post-task scores in this sample; the design does not isolate a causal effect. |
| Click counts measured engagement. | Clicks are behavioral traces and may not represent engagement. | Click counts were treated as a behavioral proxy for platform activity; engagement interpretation requires additional validity evidence. |
| The model explains student success. | Prediction does not equal explanation. | The model predicted later scores with held-out validation; explanatory mechanisms require separate evidence. |
| Students became self-regulated. | Trace patterns alone rarely prove construct change. | Trace patterns were consistent with selected SRL behaviors defined in the coding scheme. |
| This sequence shows the learning process. | A sequence is a representation shaped by event definitions and logging. | This event sequence characterizes recorded platform actions during the task. |
| The intervention worked. | "Worked" needs specified outcome and design. | The intervention was associated with favorable outcome differences under the study's design assumptions. |
| Interviews proved the pattern. | Qualitative evidence interprets or contextualizes rather than proving a quantitative trend. | Interviews helped interpret why the observed pattern may have occurred. |
| Our findings generalize to online learners. | Needs sampling, replication, or generalization evidence. | Findings are bounded to the learners, platform, and task context represented in this dataset. |