Assignment methods for incidence calculus

Incidence calculus is a mechanism for probabilistic reasoning in which sets of possible worlds, called incidences, are associated with axioms, and probabilities are then associated with these sets. Inference rules are used to deduce bounds on the incidence of formulae which are not axioms, and bounds for the probability of such a formula can then be obtained. In practice an assignment of probabilities directly to axioms may be given, and it is then necessary to find an assignment of incidence which will reproduce these probabilities. We show that this task of assigning incidences can be viewed as a tree searching problem, and two techniques for performing this research are discussed. One of these is a new proposal involving a depth first search, while the other incorporates a random element. A Prolog implementation of these methods has been developed. The two approaches are compared for efficiency and the significance of their results are discussed. Finally we discuss a new proposal for applying techniques from linear programming to incidence calculus.

Who is susceptible to conjunction fallacies in category-based induction?

Recent evidence suggests that the conjunction fallacy observed in people's probabilistic reasoning is also to be found in their evaluations of inductive argument strength. We presented 130 participants with materials likely to produce a conjunction fallacy either by virtue of a shared categorical or a causal relationship between the categories in the argument. We also took a measure of participants' cognitive ability. We observed conjunction fallacies overall with both sets of materials but found an association with ability for the categorical materials only. Our results have implications for accounts of individual differences in reasoning, for the relevance theory of induction, and for the recent claim that causal knowledge is important in inductive reasoning.

Learning about probability from text and tables: Do color coding and labeling through an interactive-user interface help?

Learning from visual representations is enhanced when learners appropriately integrate corresponding visual and verbal information. This study examined the effects of two methods of promoting integration, color coding and labeling, on learning about probabilistic reasoning from a table and text. Undergraduate students (N = 98) were randomly assigned to learn about probabilistic reasoning from one of 4 computer-based lessons generated from a 2 (color coding/no color coding) by 2 (labeling/no labeling) between-subjects design. Learners added the labels or color coding at their own pace by clicking buttons in a computer-based lesson. Participants' eye movements were recorded while viewing the lesson. Labeling was beneficial for learning, but color coding was not. In addition, labeling, but not color coding, increased attention to important information in the table and time with the lesson. Both labeling and color coding increased looks between the text and corresponding information in the table. The findings provide support for the multimedia principle, and they suggest that providing labeling enhances learning about probabilistic reasoning from text and tables

A Ligand predication tool based on modeling and reasoning with imprecise probabilistic knowledge

Ligand prediction has been driven by a fundamental desire to understand more about how biomolecules recognize their ligands and by the commercial imperative to develop new drugs. Most of the current available software systems are very complex and time-consuming to use. Therefore, developing simple and efficient tools to perform initial screening of interesting compounds is an appealing idea. In this paper, we introduce our tool for very rapid screening for likely ligands (either substrates or inhibitors) based on reasoning with imprecise probabilistic knowledge elicited from past experiments. Probabilistic knowledge is input to the system via a user-friendly interface showing a base compound structure. A prediction of whether a particular compound is a substrate is queried against the acquired probabilistic knowledge base and a probability is returned as an indication of the prediction. This tool will be particularly useful in situations where a number of similar compounds have been screened experimentally, but information is not available for all possible members of that group of compounds. We use two case studies to demonstrate how to use the tool.

An extended framework for evidential reasoning systems.

Use of the Dempster-Shafer (D-S) theory of evidence to deal with uncertainty in knowledge-based systems has been widely addressed. Several AI implementations have been undertaken based on the D-S theory of evidence or the extended theory. But the representation of uncertain relationships between evidence and hypothesis groups (heuristic knowledge) is still a major problem. This paper presents an approach to representing such knowledge, in which Yen’s probabilistic multi-set mappings have been extended to evidential mappings, and Shafer’s partition technique is used to get the mass function in a complex evidence space. Then, a new graphic method for describing the knowledge is introduced which is an extension of the graphic model by Lowrance et al. Finally, an extended framework for evidential reasoning systems is specified.

Constructing probabilistic ATMS using extended incidence calculus

This paper discusses the relations between extended incidence calculus and assumption-based truth maintenance systems (ATMSs). We first prove that managing labels for statements (nodes) in an ATMS is equivalent to producing incidence sets of these statements in extended incidence calculus. We then demonstrate that the justification set for a node is functionally equivalent to the implication relation set for the same node in extended incidence calculus. As a consequence, extended incidence calculus can provide justifications for an ATMS, because implication relation sets are discovered by the system automatically. We also show that extended incidence calculus provides a theoretical basis for constructing a probabilistic ATMS by associating proper probability distributions on assumptions. In this way, we can not only produce labels for all nodes in the system, but also calculate the probability of any of such nodes in it. The nogood environments can also be obtained automatically. Therefore, extended incidence calculus and the ATMS are equivalent in carrying out inferences at both the symbolic level and the numerical level. This extends a result due to Laskey and Lehner.

Adaptive Dialogue Strategy Selection through Imprecise Probabilistic Query Answering

In a human-computer dialogue system, the dialogue strategy can range from very restrictive to highly flexible. Each specific dialogue style has its pros and cons and a dialogue system needs to select the most appropriate style for a given user. During the course of interaction, the dialogue style can change based on a user’s response and the system observation of the user. This allows a dialogue system to understand a user better and provide a more suitable way of communication. Since measures of the quality of the user’s interaction with the system can be incomplete and uncertain, frameworks for reasoning with uncertain and incomplete information can help the system make better decisions when it chooses a dialogue strategy. In this paper, we investigate how to select a dialogue strategy based on aggregating the factors detected during the interaction with the user. For this purpose, we use probabilistic logic programming (PLP) to model probabilistic knowledge about how these factors will affect the degree of freedom of a dialogue. When a dialogue system needs to know which strategy is more suitable, an appropriate query can be executed against the PLP and a probabilistic solution with a degree of satisfaction is returned. The degree of satisfaction reveals how much the system can trust the probability attached to the solution.

When does information about causal structure improve statistical reasoning?

Base rate neglect on the mammography problem can be overcome by explicitly presenting a causal basis for the typically vague false-positive statistic. One account of this causal facilitation effect is that people make probabilistic judgements over intuitive causal models parameterized with the evidence in the problem. Poorly defined or difficult-to-map evidence interferes with this process, leading to errors in statistical reasoning. To assess whether the construction of parameterized causal representations is an intuitive or deliberative process, in Experiment 1 we combined a secondary load paradigm with manipulations of the presence or absence of an alternative cause in typical statistical reasoning problems. We found limited effects of a secondary load, no evidence that information about an alternative cause improves statistical reasoning, but some evidence that it reduces base rate neglect errors. In Experiments 2 and 3 where we did not impose a load, we observed causal facilitation effects. The amount of Bayesian responding in the causal conditions was impervious to the presence of a load (Experiment 1) and to the precise statistical information that was presented (Experiment 3). However, we found less Bayesian responding in the causal condition than previously reported. We conclude with a discussion of the implications of our findings and the suggestion that there may be population effects in the accuracy of statistical reasoning.

Whose statistical reasoning is improved by information about causal structure?

People often struggle when making Bayesian probabilistic estimates on the basis of competing sources of statistical evidence. Recently, Krynski and Tenenbaum (Journal of Experimental Psychology: General, 136, 430–450, 2007) proposed that a causal Bayesian framework accounts for peoples’ errors in Bayesian reasoning and showed that, by clarifying the causal relations among the pieces of evidence, judgments on a classic statistical reasoning problem could be significantly improved. We aimed to understand whose statistical reasoning is facilitated by the causal structure intervention. In Experiment 1, although we observed causal facilitation effects overall, the effect was confined to participants high in numeracy. We did not find an overall facilitation effect in Experiment 2 but did replicate the earlier interaction between numerical ability and the presence or absence of causal content. This effect held when we controlled for general cognitive ability and thinking disposition. Our results suggest that clarifying causal structure facilitates Bayesian judgments, but only for participants with sufficient understanding of basic concepts in probability and statistics.

Probabilistic logic with independence

This paper investigates probabilistic logics endowed with independence relations. We review propositional probabilistic languages without and with independence. We then consider graph-theoretic representations for propositional probabilistic logic with independence; complexity is analyzed, algorithms are derived, and examples are discussed. Finally, we examine a restricted first-order probabilistic logic that generalizes relational Bayesian networks.