GAFExtractor

The GAFExtractor is the driving algorithmic class of Argflow, being responsible for combining the different modular elements of the library (argumentation frameworks, mappers and Chi) and applying them to a given model with a specific input for generating an explanation.

Constructor

GAFExtractor(influence_mapper, strength_mapper, characterisation_mapper, chi)

Create a GAFExtractor to which a model can be fed in order to extract an explanation based on the mechanisms described by its arguments, which we detail below.

• influence_mapper - a function that generates the relevant influence graph from a model given an input

• strength_mapper - a function that provides a strength given the source and destination of an edge

• characterisation_mapper - a function that provides the relevant characterisation for an argument

• chi - a Chi instance that generates visualisations for an argument

Methods

extract(self, model, x)

This is a class method that takes a model and some input for it and returns a GAF, embellished with the required payloads (with an inferred PayloadType), which can then be serialized for communication with the Portal API, which will generatwe an explanation and the corresponding conversational features.

• model - a Model.

• x - some input to the Model.