The Attention Engine is a containerized psychometric monitoring framework. It utilizes a jsPsych frontend matrix to capture sub-millisecond keyboard reaction timings, streaming trial payload arrays directly to a FastAPI Python backend engine over a secure token-validated network bridge. Persistent state rows are recorded locally within an optimized SQLite data layer.
Use these standard peer-reviewed population benchmarks to audit baseline tracking datasets:
| Metric Name | Task Context | Population Baseline | Operational Significance |
|---|---|---|---|
| Sensitivity (d') | 2-Back Memory | 2.20 | SD: ±0.75 | Measures information preservation capacity inside your short-term working memory buffer. Lower values capture rapid signal decay or buffer dropouts. |
| Response Bias (c) | 2-Back Memory | 0.00 (Balanced Style) | Evaluates internal decision thresholds. Negative ranges capture impulsive velocity tactics, while positive scores signal a cautious accuracy bias. |
| Inhibition Cost | Stroop Shield | 280.0 ms | SD: ±85.0 ms | The literal cognitive reaction tax paid to suppress competing textual noise in order to map out ink targets. Faster (lower) speeds signal peak selective filtering. |
To preserve long-term analytical trends from environment irregularities (such as OS micro-stutters, background processing spikes, or erratic layout clicks), an automated processing filter guards the database gate. Incoming sessions throwing unfeasible scores (Inhibition Cost exceeding ±2500 ms, or a working memory sensitivity scaling past 4.5) are systematically dropped at the API barrier to protect platform integrity.