A naturalistic decision making model for simulated human combatants

Abstract

We describe a naturalistic behavioral model for the simulation of small unit combat. This model, Klein's recognition-primed decision ma.khg (RPD) model[l], is driven by situational awareness rather than a rational process of selecting from a set of action options. We argue that simulated combatants modeled with RFD will have more flexible and realistic responses to a broad range of small-scale combat scenarios. Furthermore, we note that the predictability of a simulation using an RPD fia.mework can be easily controlled to provide multiple evaluations of a given combat scenario. Finally, we discuss computational issues for buildlng an RPD-based behavior engine for fully automated combatants in small conflict scenarios, which are being investigated within %ndia's Next Generation Site Security project.

Controlling uncertainty: A review of human behavior in complex dynamic environments.

“…)), and feedback (delayed or immediate)). For instance, even when individuals have become highly skilled, certain objective task characteristics of a CDC task may produce highly erratic and damaging outcomes (e.g., Degani, 2004;Hunter, Hart, complexity in this way has not always been a successful predictor of performance on CDC tasks (e.g., Campbell, 1988;Quesada et al, 2005). However, direct comparison of CDC tasks on multiple dimensions of task complexity has yet to be investigated, and until work of this nature is conducted, this issue will remain.…”

“…Low uncertainty. The alternation between monitoring (task monitoring, self-monitoring) and control behaviors is less frequent when the individual judges the environment to be sufficiently familiar to draw on highly practiced plans of actions (Cohen et al, 1996;Kaber & Endsley, 2004;Goa & Lee, 2006;Hunter et al, 2000;Kirlik, 2007;Kleinmuntz, 1985;Sarter et al, 2007;Vicente et al, 2001;Yeo & Neal, 2006), or is deemed predictable enough to plan actions and decisions online (e.g., Hogarth & Makridakis, 1981;Kerstholt, 1996). Typically, people feel confident of judging accurately the outcome of their actions, and so this lowers the demand on cognitive resources (Camp et al, 2001;Jones & Mitchell, 1994;Kluge, 2008aKluge, , 2008bMosier et al, 2007;Orasanu & Connolly, 1993;Vicente, 2002;Yeo & Neal, 2006), because a highly familiar series of planned actions or decisions, which do not do require continual tracking of action-outcome, are implemented (e.g., Metzger & Parasuraman, 2005).…”

Controlling Uncertainty

“…These biases, while offering a basis on which to test hypotheses and develop plans of actions that can advance the individual towards a goal, interfere with forming accurate representations of the environment, and prolonged reliance on biases leads to poor task knowledge, and poor control. As with biases, pattern-matching is a speedy means of isolating any features of the task that may relate to previous experiences of similar situations (Berry & Broadbent, 1984Degani, et al, 2006;DeShon & Alexander, 1996;Hunter, et al, 2000;Schmitt & Klein, 1996), and these can be used as a basis for developing appropriate strategies (e.g., Kobus, Proctor, & Holste, 2001). However, evoking previously developed plans of actions through pattern matching may also lead to ineffective strategy application, and if such plans are inappropriate for the current CDC task then inaccurate knowledge of the task develops (Brandouy, 2001;Brézillon, et al, 1998;Dienes & Fahey, 1995;Geddes & Stevenson, 1997;Diehl & Sterman, 1995;Monxes (2000).…”

Naturalistic Decision Making in Power Grid Operations: Implications for Dispatcher Training and Usability Testing

“…Following the August 14, 2003 blackout, the RPDM was introduced to the power industry and was successfully applied to the planning, development, and implementation of grid operator training systems such as the Virtual Instructor and PowerSimulator (Podmore et al, 2008). A descriptive diagram of the RPDM is shown in Figure 1 (after Hunter et al, 2000). This figure shows that while experts may typically immediately recognize the situation and see the applicability of associated actions (following the path of the unshaded process in Figure 1), in some cases-such as novel/unfamiliar situations-the cues in the environment do not immediately match known or expected patterns and problems.…”