Building a Knowledge Base to Support the Authoring of Social Skills Instructional Modules

Background:

Difficulties with social skills are generally considered defining characteristics of HFA. Because deficits in socialization interfere with the educational experiences and quality of life of individuals with HFA, and because interventions must be highly individualized to be effective, we are interested in exploring the way technologies may play a role in assisting in the creation of customized social skills instructional modules.

Objectives:

Last year at IMFAR we presented Refl-ex, a computer-based system designed to allow the individual with autism to practice social problem solving skills by experiencing social situations and choosing appropriate responses to unexpected events. It is our goal to develop a system that will help non-experts (i.e. authors who have little or no knowledge of instructional strategies) author and share Refl-ex instructional modules. We call the authoring tool Refl-ex Authoring and Critiquing Tool (REACT).

Our first steps will be to build up the knowledge base of the authoring tool, by collecting information about individuals’ cognitive scripts for every day tasks using crowd sourcing. A cognitive script is a standard event sequences we develop, which enables us to subconsciously know what to expect in particular situations. Crowd-sourcing is the process of delegating a particular task to a distributed group of people. Through this process, we will build a model that shows all the ways a situation can unfold, and possible obstacles that can arise.

Methods:

We have conducted a study in which we use crowd-sourcing strategies to elicit individuals’ cognitive scripts for everyday events. Namely going to a restaurant and going to a movie. In addition, we asked participants what could go wrong at each step. The data collected is being analyzed to develop a model, which is a graph showing probabilistically how events follow each other; this model shows all the ways in which a restaurant experience can unfold. The introduction of an obstacle is salient to our pedagogical approach. For this reason our model will also enable REACT to provide the author with ideas for obstacles and possible solutions to insert into the modules. This model will be used to build the knowledge base and provide suggestions.

Results:

The data collected from the cognitive scripts study has generated interesting results with respect to the regularity of the steps taken in a script describing an everyday task, despite the diversity in the language used. In addition, collecting large amounts of data makes it possible to create the graph of probabilities and provide diverse suggestions for language to describe a particular step.

Conclusions:

Crowd-sourcing techniques makes it possible to collect data that can be used to model knowledge in the world. In our study, we elicited participant’s scripts for everyday tasks. This data has enabled us to create a model that can be used to aid authors in the creation of the instructional modules. By providing this information, and enabling collaboration between the author and the system, we believe it will be possible to create highly customized social skills instructional modules.