next up previous contents
Next: AVAILABLE TECHNOLOGIES Up: PRAGMATICS Previous: Discourse Agents   Contents

Logical Abduction

The normal, sound, use of modus ponens is to infer that \( \beta \) is true given that \( \alpha \) and \( \alpha \Rightarrow \beta \) are known. However Figure 2.12 shows the type of unsound reasoning known as logical abduction.

Figure 2.12: Logical abduction
\begin{figure}\begin{displaymath}
\frac{\begin{array}{l}
\alpha \Rightarrow \beta \\
\beta
\end{array}}{\alpha }\end{displaymath}\par\end{figure}

For example with the facts ``Jeff can fly'' and ``ducks can fly'', logical abduction allows the conclusion ``Jeff is a duck'' to be inferred, which could obviously be incorrect.

Generally an effect \( \beta \) can have many \( \alpha _{i} \) causes, and a method to choose the 'best' abductive proof is needed. There exist several techniques for choosing between various explanations, including probabilistic methods (which have the difficulty of defining the space over which to calculate the probabilities, and calculate them given the lack of training data) and heuristic methods which use another evaluation method such as choosing the explanation with the least number of assumptions. These methods are considered beyond the scope of this paper and shall therefore not be discussed.



Andrew P Coates (UG) 2002-07-17