Adaptive Agents for Fit-for-Purpose Training

Adaptive Agents for Fit-for-Purpose Training

van den Bosch, K.
Blankendaal, R.A.M.
Boonekamp, R.
Schoonderwoerd, T.A.J.

Stephanidis, C. (editor)
Harris, D. (editor)
Li, W.C. (editor)
Schmorrow, D.D. (editor)
Fidopiastis, C.M. (editor)
Zaphiris, P. (editor)
Ioannou, A. (editor)
Fang, X. (editor)
Sottilare, R.A. (editor)
Schwarz, J. (editor)

2020

Simulators and games provide contextually rich environments, enabling learners to experience the relations between actions, events and outcomes. In order to be effective, learning situations need to be tailored to the needs of the individual learner. Virtual characters (or agents) that, in real time, select, adapt, and exhibit the behavior that is exactly right for that learner, help to establish such fit-for-purpose training. This paper discusses principles for designing training with adaptive agents, and presents a framework for their autonomous and dynamic operation. A prerequisite for agents’ adaptation of behavior to be successful is that adjustments do not violate the consistency and believability of the character, and maintains the overall narrative of the scenario. For reasons of management and coordination, it is proposed not to assign control over adaptations to virtual character-agents themselves, but to a dedicated director agent. This director agent is not a virtual character in the gameplay, but operates in the background. It collects and manages information, makes decisions about adaptations and issues behavioral instructions to the virtual characters agents. The framework was used in a pilot study, employing a human facilitator that simulated a director agent, arranging the adaptive behavior of virtual characters in a game-based training of military tactical decision making. Effects of adaptive and non-adaptive agents in a training were compared. Adaptive agents had a positive influence on learning and performance, and an increased engagement and appreciation by learners. Additional research with more participants is needed to verify these preliminary findings. © 2020, Springer Nature Switzerland AG.

Adaptive agent behavior
Adaptive instructional systems
Cognitive behavior
Director agent
Intelligent agents
Learner modeling
Learning
Personalized learning
Serious games
Simulation
Training
Behavioral research
Decision makingHuman computer interaction
Adaptive agents
Adaptive behavior
Dynamic operations
Fit for purpose
Game-based trainings
Learning situation
Tactical decision makings
Virtual character
Autonomous agents

http://resolver.tudelft.nl/uuid:6aa5ee7a-92b3-4bfb-b001-473e00a617e3

https://doi.org/10.1007/978-3-030-60128-7_43

882240

Springer, Cham

9783030601287

HCI International 2020 – Late Breaking Papers: Cognition, Learning and Games. HCII 2020. Lecture Notes in Computer Science, 12425 (12425), 586-604

conference paper