CoJACK is a novel BDI cognitive architecture that uses psychological principles to predict how humans will behave in a given situation.
In addition to its procedural knowledge (“knowing how to do something”), CoJACK agents have declarative knowledge (“knowing about”). In the “Sudden enemy fire” diagram above, knowing that the squad leader has been killed is an example of declarative knowledge. Declarative knowledge is stored in beliefsets within the agent.
The mental process of selecting an action to perform is termed decision making.
Many factors can contribute to a decision, including:
1. Situation Awareness – the appreciation of those aspects of the current situation that are relevant to the question at hand.
2. Predictive Capability – the ability of the agent to foresee the consequences of actions and the likely actions/reactions of other entities that are part of the scenario.
3. Response Repertoire – the known action sequences for dealing with the current situation (skill set).
4. Personal Preference – preferred methods of dealing with the current situation, often based on experience of previous successes and failures.
5. Cognitive Effectiveness – the current state of the underlying cognitive architecture, affecting capabilities such as ability to recall facts, hold intermediate results in working memory, and stay focused on the problem.
6. Affective State – the emotional factors that can influence a decision, for example, a high-level of fear can predispose a person to make an irrational decision.
These factors contribute to individual differences in decision making and, ideally, should be reflected in a synthetic human agent. Typically, synthetic human entities are implemented in relatively inflexible scripting languages. In contrast, JACK® offers a high-level plan language that provides intelligent and flexible responses to the situation. Using JACK, one can represent the first four of the above decision making factors. However, a JACK agent will cooly execute its plans without error and as fast as the hardware platform allows. As such, it fails to model the underlying properties of and variability in human cognition; a JACK agent will not forget what it knows and can perform computational feats that are well outside the capabilities of humans. This leads to unrealistic performance because even highly-trained humans vary in their response to a given situation. This, then, was the motivation for adding a cognitive architecture to JACK. CoJACK alters the performance of a JACK agent so that it more closely resembles how a human would perform the task at hand.