Activity states framework (ASF)

What?

A theoretical framework that experiments with the notion of - conceptualization, and contextualization from situated cognition, and psychic reflection from activity theory for studying, and modeling context into a method called the activity states framework (ASF).

Why?

To provide a method of analysis for identifying collaborators activity during situated context. We foresee that the results of the work can be used to guide human-computer interaction design for Web-mediated collaborative dialogs, and for understanding better the notion of conceptualization.

How?

Using a combination of situated cognition, and activity theory to guide the design of the computer model. We use the notion of conceptualization, contextualization, and activity theory from situated cognition together with the notion of reflection from activity theory to guide the design of the computer model.

Where are we now?

We use situated cognition, and activity theory as the underpinning theoretical foundation to develop a method for analyzing collaborative dialogs, known as the 'activity states framework'.

What is 'Activity States' Framework?

The term 'activity states' is based on situated cognition (Clancey, 1997), and activity theory (Leont'ev, 1978). It is defined as the level of attention that a speaker is engaged in during his construction of his situated activity, as illustrated in Figure 1 below:

Figure 1. Collaborator's level of attention.

Refer to Figure 1. The red line represents the level of attention of a collaborator – that is categorized from passive, semi-active, to active (see the text in blue). Passive signi- fies that a collaborator is just about to begin a new activity. Semi-active signifies that a collaborator is pursuing the activity. While active signifies that the collaborator is about to reach the object of the activity. The squares represent the type of activities that collaborators are engaged in. Overlapping squares represent that the collaborators are multi-tasking their activities during the collaboration.

The ASF workflow

The ASF workflow provides steps on how to analyze the dialogs for identifying collaborators situated activity based on the basic idea of ASF. We show below the ASF workflow.

Figure 2. ASF workflow.

We implemented the workflow into a computer program (a system). For more details on the ASF workflow, click here for the recently published paper.

At the moment, we have tested the system with naturalistic chat dialogs, and actual co-located meetings to identify context of 'conflict', i.e., disagree. The system at the moment can identify the speech acts of each utterances. This was necessary in order to identify the activity (i.e., context) of the collaborators.

An example of the kind of dialogs that the system can analyze, we show an excerpt of an actual chat between two collaborators in Figure 3 below.

Figure 3. Sample of an input fille.

Now we will show the screenshots of the activity states system.

Figure 4. The activity states user interface.

The dialogs that has been analyzed will be plotted accordingly to the 'activity states', passive, semi-active or active on a timeline.

Figure 5. Click on icon labeled 'Analyze', then select the input file of dialogs that you would like to analyze.

When the label 'Analyze' is clicked, you will be prompted by the open input file.

Figure 6. Each utterance analyzed plotted to its correspondingly activity states.

Each utterance will be plotted into its correspondingly analyzed activity states, whether the utterance 'well that's the kind of q..' indicates that the speaker's level of attention is in a 'semi-active state' based on the analyzed speech acts.

Figure 7. Red indicates context of 'conflict'.

Since the system at the moment is tested to identify context of conflict, we mark the context with the color red. One may zoom into the red area to have a detailed view of the analyzed utterance.

Figure 8. Zoom in for the details of the analyzed utterance.

The analyzed utterances are marked up into agent messages in the following format: speaker speech acts content. For more details, read our paper here