Adding in-game strategy features to pre-test features substantially improves LSTM-based stealth assessment accuracy¶

Claim · All claims
Evidence · 1 study · design r2 · q2
Currency · newest study 2017 · not yet checked against its sources · not reviewed by a person · overdue for a recheck since 2020: tool and model accuracy changes quickly

Subclaims¶

q2 i? The LSTM model using pre-test plus in-game strategy features achieved 64.82% accuracy versus 44.66% for the pre-test-only version. → Akram 2017

Evidence¶

Akram 2017¶

Akram, B., Mott, B., Min, W., Boyer, K. E., Wiebe, E., & Lester, J. (2017). Improving Stealth Assessment in Game-based Learning with LSTM-based Analytics. https://scholar.google.com/scholar?q=Improving+Stealth+Assessment+in+Game-based+Learning+with+LSTM-based+Analytics

q2 · i? · design · r2

Comparison of two LSTM versions on the same train/test split of 168 students' data: full feature set (pre-test plus strategy cluster sequences) versus pre-test-only feature set. The full-set model reached "an accuracy of 64.82%" against 44.66% for pre-test only (Table 3).

"Compared to the 44.66% accuracy achieved by the partial feature set version (pre-test features only), the model that u ses in-game strategy features in addition to pre -test features achieves an accuracy of 64.82%."

Discussion¶

Related Claims¶

• An LSTM-based strategy temporal analytics framework outperforms random forest, SVM, and majority-class baselines for stealth assessment predictive accuracy
• Clustering n-gram flip sequences per task yields three distinct problem-solving strategy groups across the binary challenges
• Stealth assessment in AI-GBL achieves accurate prediction of learner states, with bias auditing demonstrated without accuracy loss
• GRU is the lowest-performing model across both single-task and multi-task competency modeling, likely due to its two gating units versus LSTMs' three — related