Mixing Automatic and Deliberative Learning During Problem Solving

Mixing Automatic and Deliberative Learning During Problem Solving Randolph M. Jones Soar Technology & Colby College March 20, 2004 Symposium on Reasoning and Learning in Cognitive Systems

Background • There alternative ways we might incorporate multi-step learning into a model – One approach would be to automate explicit instruction of desired task behavior • Even this is difficult – This talk focuses on models that can discover new problem-solving knowledge and strategies on their own March 20, 2004 Mixing Automatic and Deliberative Learning During Problem Solving 2

Knowledge Tuning and Acquisition • There are two primary ways a model can learn new strategies – Acquiring new task knowledge that allows more complete or efficient coverage of a problem space – Tuning existing task knowledge so it is retrieved more oportunistically • Knowledge acquisition in its own right is also important – But this work suggests that knowledge acquisition depends on knowledge tuning March 20, 2004 Mixing Automatic and Deliberative Learning During Problem Solving 3

Knowledge Tuning • Basic representational structure of knowledge “chunk” remains unchanged • Retrieval/selection patterns associated with the knowledge do change March 20, 2004 Mixing Automatic and Deliberative Learning During Problem Solving 4

Knowledge Acquisition • Entirely new structured representations of long-term knowledge are added to the model’s knowledge base • Or existing “chunks” of knowledge undergo structural changes March 20, 2004 Mixing Automatic and Deliberative Learning During Problem Solving 5

Task Example: Solving Physics Problems • Learning to solve physics problems involves learning new equations relevant to the problems, and learning the situations in which those equations should be used • Students who “self-explain” study examples show greater improved performance than those who don’t (Chi et al. , 1989) – Are they tuning knowledge or acquiring knowledge? • Cascade (Van. Lehn, Jones, & Chi, 1992) models the “selfexplanation effect” observed in humans learning to solve physics problems March 20, 2004 Mixing Automatic and Deliberative Learning During Problem Solving 6

Task Example: Simple Addition • There a variety of strategies that can be used to perform elementary addition, some more efficient than others • Children are usually instructed using a basic strategy, but invent a particular set of more efficient strategies on their own (Siegler & Jenkins, 1989) – Are they tuning knowledge or acquiring knowledge? • GIPS (Jones & Van. Lehn, 1994) models the series of strategy shifts exhibited by children March 20, 2004 Mixing Automatic and Deliberative Learning During Problem Solving 7

Cascade: Typical Problem 10 kg What is the tension in the string? March 20, 2004 Mixing Automatic and Deliberative Learning During Problem Solving 8

Cascade: Typical Problem A B C What is the magnitude of each force? March 20, 2004 Mixing Automatic and Deliberative Learning During Problem Solving 9

Cascade: Typical Example • Let the knot be “the body” • FA, FB, FC are all the forces acting on the body • The body is at rest, so FA+FB+FC=0 • By projection, FAX+FBX=0 • By projection, FAY+FBY+FCY=0 • FAX=–FA cos 30 = – 0. 8666 FA • etc. March 20, 2004 FB FA Mixing Automatic and Deliberative Learning During Problem Solving FC 10

Cascade: Modeling Goal • Explain the learning process and other factors that cause students who carefully study examples to learn more effectively than students who do not March 20, 2004 Mixing Automatic and Deliberative Learning During Problem Solving 11

Cascade: Knowledge Representation • Long-term task knowledge is a set of physics equations, geometric equations, and rules for representing free-body diagrams – Implemented in Prolog • Default problem-solving strategy is exhaustive depth-first search with backtracking – Straightforward application of Prolog • Problem-solving goals are quantities (variables) for which the problem solver must compute a value • Selection knowledge allows heuristic search by using past solution paths as analogies to the current problem March 20, 2004 Mixing Automatic and Deliberative Learning During Problem Solving 12

Cascade: Learning Processes • Knowledge tuning – Analogical Search Control – When Cascade succeeds in computing a value for a sought quantity, it records a triple including the name of the problem, the sought quantity, and the equation that was used to compute the value – The caching process occurs automatically and frequently, every time a subgoal is achieved – On subsequent problems, Cascade • Attempts to map the current problem quantities and relations to the analog problem • Searches for cached triples that mention problem analogs to the current problem, together with an analogous sought quantity • Attempts the retrieved equation before falling back on the default ordering of knowledge (if backtracking occurs) March 20, 2004 Mixing Automatic and Deliberative Learning During Problem Solving 13

Cascade: Learning Processes • Knowledge acquisition – Explanation-based Learning of Correctness – If Cascade cannot solve a problem (after exhaustive search), it begins the search again, this time attempting a “repair” at the first point that backtracking is encountered • Repairs occur by attempting to apply relevant “overly general rules” to the problem • On success, Cascade stores a specialization of the overly general rule with the rest of the task knowledge – The rule learning process occurs deliberatively and infrequently, only after the model has recognized an impasse in problem solving March 20, 2004 Mixing Automatic and Deliberative Learning During Problem Solving 14

Cascade: Learning Interactions • Knowledge acquisition only works if the model is repairing “the right gap” in a potential solution space • The model can be guided toward the right gap: – By the directions in a worked example – By the quality of knowledge tuning March 20, 2004 Mixing Automatic and Deliberative Learning During Problem Solving 15

Cascade: Learning Interactions Initial problem state False paths Dead ends Knowledge gap Solution March 20, 2004 Mixing Automatic and Deliberative Learning During Problem Solving 16

Cascade: Experimental Results • No Analogical Search Control – Learns 3 correct rules – Solves 9 problems correctly • No EBLC on examples – Learns 13 correct rules – Learns 4 incorrect rules – Solves 21 problems correctly (many using a backup transformational analogy strategy) • ASC & EBLC – Learns 22 correct rules – Solves 23 problems correctly March 20, 2004 Mixing Automatic and Deliberative Learning During Problem Solving 17

GIPS: Typical Problem “Sum” Strategy March 20, 2004 Mixing Automatic and Deliberative Learning During Problem Solving 18

GIPS: Modeling Goal • Model how children independently invent the “Min” strategy with experience – “Min” is a more efficient strategy, suggesting that it may be produced primarily by knowledge tuning – However, there appear to be structural changes to the steps the children are taking to solve problems March 20, 2004 Mixing Automatic and Deliberative Learning During Problem Solving 19

GIPS: Knowledge Representation • Task knowledge is represented as STRIPS-like operators with preconditions, constraints, add conditions, and delete conditions – Problem-solving algorithm is “flexible” means-ends analysis • TRANSFORM goal: Use features describing current state and goal to retrieve a candidate operator to APPLY for the next step in the transformation • APPLY goal: Execute the operator if possible, else set up a new TRANSFORM to the preconditions of the operator • Retrieval/selection knowledge is encoded as probability estimates (for logical sufficiency and logical necessity) attached to each potential triggering feature for each operator – State and Goal relations March 20, 2004 Mixing Automatic and Deliberative Learning During Problem Solving 20

Example Bayesian Concept • “Liftable” FEATURE LS LN size is small 3. 0 0. 0 weight is light 2. 0 0. 3 has handle 2. 0 0. 3 attached to floor 0. 0 3. 0 color is red 1. 0 • Note this example has propositional features, but features in GIPS are relational – GIPS uses a graph-based “maximal partial match” procedure to map combinations of relations to “propositions” March 20, 2004 Mixing Automatic and Deliberative Learning During Problem Solving 21

GIPS: Learning Processes • Knowledge tuning – Every time an APPLY goal leads to success or failure, GIPS updates the appropriate probability estimates for each state and goal feature present when the APPLY goal was created • A = “Action A is the right thing to do next” • F = “Feature F is true in the problem situation” – A similar process occurs every time an operator executes (or not) March 20, 2004 Mixing Automatic and Deliberative Learning During Problem Solving 22

GIPS: Learning Processes • Knowledge acquisition – When feature values for an operator’s “execution concept” receive particularly strong “logical necessity” values, a deliberative process explicitly adds the new feature as a condition of the operator • Another process removes features from the operator conditions March 20, 2004 Mixing Automatic and Deliberative Learning During Problem Solving 23

The SUM-to-MIN Strategy Shift I see X fingers and X is the addend. (I don’t have to count! I can use the counter for something else!) I just counted “X” and X is the addend. (I have to count. ) March 20, 2004 X is the addend. (I don’t have to raise any fingers. ) Mixing Automatic and Deliberative Learning During Problem Solving 24

GIPS Learning Interactions • Bayesian updates happen continuously and automatically, leading to performance shifts based on retrieval of operators • Based on accumulating evidence, the model periodically tries more drastic structural changes to operator preconditions, which have larger effects on subsequent retrieval patterns (because operator preconditions are used as subgoal retrieval cues and determine satisfaction of APPLY goals) March 20, 2004 Mixing Automatic and Deliberative Learning During Problem Solving 25

Lessons • It is difficult to acquire new knowledge without first tuning old knowledge • Tuning old knowledge implies that you have some old knowledge to tune • For complex learning, we need to focus on learning in the context of significant prior knowledge • Tuning can help guide the search for building new operators (Cascade) as well as for adjusting the structural representations of existing operators (GIPS) – You only want to acquire new knowledge after you have accumulated some evidence (from tuning) that the knew knowledge is appropriate and useful March 20, 2004 Mixing Automatic and Deliberative Learning During Problem Solving 26