For handing out at Cognition 2006, Montreal Lessons from decompiling an embodied cognitive model
Bibliographic record
Abstract
Cognitive models and intelligent agents are becoming more complex and pervasive. It is time again to consider high-level behavior representation languages and development environments that make it easier to create, share, and reuse cognitive models. One of these languages is Herbal, a high-level behavior representation language. Users represent knowledge in Protégé, an ontology editor. Herbal compiles this knowledge into cognitive models in Soar, a rule-based cognitive architecture. Herbal includes the ability to automatically link the resulting models to dTank, a simple, distributed tank game co-developed with Herbal. To understand the theoretical implications of the process of compiling cognitive models from high-level descriptions more clearly, we generated by hand the Herbal high-level description of a well-written, medium-sized (50 rule) Soar model that plays dTank. This process is a type of decompilation process, of going from low- to high-level language, that yields lessons for both the compiler and the process of modeling. Many of the constructs in the model were supported by Herbal, particularly elaborations and simple and regular actions. In some cases, Herbal’s representation prevents the user from generating incomplete, incorrect or atheoretical code—we saw hand written code that cannot be generated by Herbal because it is incorrect or overgeneral. This process also highlights problems with Herbal. Certain types of theoretically sound handwritten rules do not yet possess an exact translation to the high-level language (mostly knowledge about which action to chose, and links across known representations). We have several suggestions for constructs to be added. This process of decompilation illustrates how users are creating models and could do so more easily and less error prone with more appropriate languages, in addition to helping develop Herbal, and, if automated, this decompilation process done by hand could lead to a decompilation feature in Herbal to help explain raw Soar code.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.004 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.003 | 0.005 |
| Scholarly communication | 0.006 | 0.009 |
| Open science | 0.001 | 0.002 |
| Research integrity | 0.002 | 0.004 |
| Insufficient payload (model declined to judge) | 0.066 | 0.009 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".