1. Chapter 13 Inference Techniques 917807 Allen

2. Konica Automates A Help Desk with Case-based Reasoning The situation to be diagnosed is entered into the system. The text is analyzed by a natural language processor to interpret the situation and compare it to existing cases stored in a case base. The user can either go with the recommended course of action or refine the wording to tighten up the description of the problem. The system can apply AI technology in solving real problems.

3. Reasoning in Artificial Intelligence A computer program is needed to access the knowledge for making inferences. This program is an algorithm that controls a reasoning process and usually called the inference engine. The inference engine directs the search through the knowledge base, a process that may involve the application of inference rules in what is called pattern matching. The most popular inference engines: forward and backward chaining.

4. The ways of people reasoning Formal reasoning methods Heuristic reasoning Common sense applied to specific goals Dividing complex problems into subproblems Parallelism-neural processors operating in parallel Analogy, or the ability to associate and relate concepts Synergy, in which the whole is greater than the sum of its parts Serendipity, or fortuitous accidents

5. Inference Methods Deductive reasoning Move from a general principle to a specific inference. Inductive reasoning Move from some established facts to draw general conclusions. Analogical reasoning Derive an answer to a question by known analogy. It is a verbalization of internalized learning process. Use of similar past experiences.

6. Inference Methods (Conti.) Formal reasoning Syntactic manipulation of a data structure to deduce new facts following prescribed rules of inferences. Procedural reasoning Use of mathematical models or simulation. Metalevel reasoning Knowledge about what is known.

7. Reasoning with Logic The most important method is called modus ponens. If A, then B In the terminology of logic, we express this as [A AND (A B ) ] B

8. Forward and Backward Chaining: An Overview There are two methods for controlling inference in rule-based ES: forward chaining and backward chaining. Example 1, pp 513 Example 2, pp513

9. Backward Chaining If the current goal is to determine the correct conclusion, then the process attempts to determine whether the premise clauses (facts) match the situation. Backward chaining is a goal-driven approach in which you start from an expectation of what is going to happen and then seek evidence that supports your expectation.

10. The Step of Backward Chaining The program starts with a goal to be verified. Look for a rule that has this goal in its conclusion. Check the premise of the rule in an attempt to satisfy the rule. It examines the assertion base first. If the search fails, the program looks for another rule. An attempt is then made to satisfy the second rule. The process continues until all the possibilities that apply are checked or until the rule initially checked is satisfied. Example 3, pp514.

11. Forward Chaining If the premise clauses match the situation, then the process attempts to assert the conclusion. Forward chaining is a data-driven approach. Start from available information, and then try to draw conclusions. The ES analyzes the problem by looking for the facts that match the IF part of its IT-THEN rules. Example 4, pp516.

12. The Inference Tree An inference tree provides a schematic view of the inference process. In building an inference tree, the premises and conclusions are shown as nodes. The branches connect the premises and the conclusions. The operators AND and OR are used to reflect the structures of the rules. Provide better insight into the structure of the rules. Visualize the process of inference and movement along its branches. It provides a guide for answering the why and how questions in the explanation process.

13. Inferencing with Frames Reasoning with frames is much more complicated than reasoning with rules. The slot provides a mechanism for a kind of reasoning called expectation-driven processing. Empty slots can be filled with data that confirm expectations. It is easy to make inferences about new objects, events, or situations because the frames provide a knowledge base drawn from previous experience.

14. Model Reasoning It is based on knowledge of the structure and behavior of the devices a system is designed to understand. Model-based systems are useful in diagnosing difficult equipment problems. It includes a model of the device to be diagnosed which is then used to identify the causes of the equipment ’ s failure.

15. Case-based reasoning Adapt solutions used to solve old problems for new problems Variation - Rule-induction method A different process of case-based reasoning - Finds cases that contain solved problems similar to the current problem - Adapts the previous solution or solutions to fit the current problem, while considering any difference between the two situations

16. Finding Relevant Cases Involves: Characterizing the input problem, by assigning appropriate features to it Retrieving the cases with those features Picking the case(s) that best match the input best

17. What is a Case? Case - Defines a problem in natural language descriptions and answers to questions, and associates with each situation a proper business action Scripts - Describe a well-known sequence of events Often “reasoning is applying scripts” Often “reasoning is applying scripts” More Scripts, Less (Real) Thinking Can be constructed from historical cases Case-based reasoning is the essence of how people reason from experience CBR - a more psychologically plausible expert reasoning model than a rule-based model

18. Advantages of Case-based Reasoning Knowledge acquisition is improved: easier to build, simpler to maintain, less expensive to develop and support. System development time is faster. Existing data and knowledge are leveraged. Complete formalized domain knowledge is not required. Experts feel better discussing concrete cases. Explanation becomes easier. Rather than showing many rules. Acquisition of new cases is easy.

19. Process of Case-based Reasoning Assign Indexes. Features of the new event are assigned as indexes characterizing the event. Retrieve. The indexes are used to retrieve a similar past case from memory. The past case contains the prior solution. Modify. The old solution is modified to conform to the new situation. Test. The proposed solution is tried out. Assign and Store. If the solution succeeds, then assign indexes and store a working solution. Explain, Repair and Test. If the solution fails, then explain the failure, repair the working solution, and test again.

20. Types of Knowledge structures Indexing rules Case memory Similarity metrics Modification rules Repair rules

21. Success Factors for a Case-base Reasoning System Determine specific business objectives. Understand your end users and customers Design the system appropriately Plan an ongoing knowledge management process Establish achievable returns on investment Plan and execute customer access strategy Expand knowledge generation and access across the enterprise

22. Explanation and Metaknowledge Explanation Human experts justify and explain their actions ES should also do so Explanation: attempt by an ES to clarify reasoning, recommendations, other actions (asking a question) Explanation facility (justifier)

23. Explanation Purposes Make the system more intelligible Uncover shortcomings of the rules and knowledge base (debugging) Explain unanticipated situations Satisfy users’ psychological and/or social needs Clarify the assumptions underlying the system's operations Conduct sensitivity analyses

24. Two Basic Explanations Why Explanations - Why is a fact requested? How Explanations - To determine how a certain conclusion or recommendation was reached.

25. Metaknowledge Knowledge about how the system reasons Knowledge about knowledge Inference rules are a special case Metaknowledge allows the system to examine the operation of the declarative and procedural knowledge in the knowledge base Explanation can be viewed as another aspect of metaknowledge Over time, metaknowledge will allow ES to create the rationale behind individual rules by reasoning from first principles

26. Generating Explanations Static Explanation: be anticipated in advance for all questions and answers. Dynamic Explanation: reconstruct explanation according to the execution pattern of the rules

27. Typology of ES Explanations Trace, or Line of Reasoning – a record of the inferential steps taken by an ES to reach a conclusion Justification - explicit description of the causal argument or rationale behind each inferential step taken by the ES

28. Inferencing with Uncertainty In Step 1, An expert provides inexact knowledge in terms of rules with likelihood values In Step 2, The inexact knowledge of the basic set of events can be directly used to draw inferences in simple cases (Step 3) In Step 3, Working with the inference engine, experts can adjust the Step 1 input after viewing the results in Steps 2 and 3.

29. Representing Uncertainty Numeric Graphic and Influence Diagram Symbolic

30. Numeric Uncertainty Representation Scale (0-1, 0-100) 0 = Complete uncertainty 1 or 100 = Complete certainty Problems with Cognitive Biases People May be Inconsistent at Different Times

31. Graphic and Influence Diagram Horizontal bars Not as accurate as numbers Experts may not have experience in marking graphic scales Many experts prefer ranking over graphic or numeric methods

32. Symbolic Uncertainty Representation Likert Scale Approach Ranking Ordinal Cardinal Pair-wise Comparison Fuzzy logic includes a special symbolic representation combined with numbers

33. Probabilities and Related Approaches The Probability Ratio The degree of confidence in a conclusion can be expressed as a probability. P(X) = Number of outcomes favoring the occurrence of X / Total number of outcomes

34. The Bayesian Extension Bayes' Theorem for combining new and existent evidence usually given as subjective probabilities a subjective probability is provided for each proposition To revise existing prior probabilities based on new information

35. Dempster-Shafer Theory of Evidence Distinguishes between uncertainty and ignorance by creating belief functions Especially appropriate for combining expert opinions, since experts do differ in their opinions with a certain degree of ignorance Assumes that the sources of information to be combined are statistically independent

36. Theory of Certainty (Certainty Factors) Uncertainty is represented as a Degree of Belief Express the Measure of Belief Manipulate degrees of belief while using knowledge- based systems Certainty Theory uses Certainty Factors Certainty Factors (CF) express belief in an event (or fact or hypothesis) based on evidence

37. Several methods of using certainty factors in handling uncertainty in knowledge-based systems 1.0 or 100 = absolute truth (complete confidence) 0 = certain falsehood CFs are NOT probabilities CFs need not sum to 100

38. Belief and Disbelief CF[P,E ] = MB[P,E] - MD[P,E] where CF = certainty factor MB = measure of belief MD = measure of disbelief P = probability E = evidence or event

39. Combining Certainty Factors (AND) IF inflation is high, CF = 50 percent, (A), AND IF unemployment rate is above 7 percent, CF =70 percent, (B), AND IF bond prices decline, CF = 100 percent, (C) THEN stock prices decline CF(A, B, and C) = Minimum[CF(A), CF(B), CF(C)] The CF for “stock prices to decline” = 50 percent The chain is as strong as its weakest link

40. Combining Certainty Factors (OR) IF inflation is low, CF = 70 percent; OR IF bond prices are high, CF = 85 percent; THEN stock prices will be high Only one IF need be true Conclusion has a CF with the maximum of the two CF (A or B) = Maximum [CF (A), CF (B)] CF = 85 percent for stock prices to be high

41. Approximate Reasoning Using Fuzzy Logic Fuzzy logic deals with quantifying and reasoning using imprecise and uncertain values Fuzzy aims at formalizing approximate reasoning process Fuzzy rules define the mapping of input variables with precise or imprecise values to output variables with precise values Fuzzy reasoning is the process of making logical inferences based on fuzzy rules and inputs