The RPD Model: Criticisms and Confusions
Six challenges to the Recognition-Primed Decision (RPD) model.
Posted February 9, 2021 | Reviewed by Ekua Hagan
Is it time to retire the Recognition-Primed Decision (RPD) model? The model initially attracted little attention (Klein, Calderwood & Clinton-Cirocco, 1986). Only when I published Sources of Power (Klein, 1998) did the RPD model come onto peoples’ radar — that was 20 years ago, long enough for it to have become outdated.
Yet the model keeps chugging along, gaining general acceptance as the way people actually make decisions. I am grateful and surprised by the RPD model’s longevity and acclaim; I hope this post doesn’t provoke a backlash.
My goal here is simply to correct some minor confusions and misrepresentations that have emerged over the decades.
The RPD model explains how fireground commanders can make good decisions within seconds. Researchers at the time thought that effective decisions depended on generating a set of options and then comparing them on evaluation dimensions. But what if you don’t have much time, or if an uncertain situation prevents careful evaluation? The RPD model shows how experienced decision-makers can do a good job even with minimal time.
It combines two processes: pattern matching to match the current situation to ones encountered in the past, which identifies reasonable courses of action, and then an evaluation process using mental simulation to imagine how that option would play out and to see if it is going to work. Thus, people can make rapid decisions without generating alternative courses of action, and without doing any comparisons.
Misunderstandings of the RPD model
1. It is a model of gut feelings, as opposed to analysis. This misunderstanding is unfortunately common and ignores the fact that the RPD model has two components: a fast, non-conscious, intuitive pattern-matching and a slower, deliberate, conscious mental simulation to do the analysis/evaluation. These two components map well to the System 1/System 2 account (e.g., Kahneman, 2011).
2. It’s easy to build a computer simulation of the RPD model. It’s only easy if you ignore the mental simulation portion and just address the pattern matching part. Even then, Artificial Intelligence simulations typically rely on reinforcement learning, strengthening and weakening connections based on their success. However, people don’t just weaken connections. We diagnose what went wrong in order to build stronger mental models.
3. The RPD model only applies to time-pressured situations. Initially, I thought this might be the case, but I found that people use the RPD tactic even in slower-paced conditions.
4. The RPD model isn’t scientific — it isn’t testable or falsifiable. I have tested it by studying inexperienced decision-makers. Predictably, they use the RPD tactic less than half the time. I also studied experienced chess players (Klein et al., 1995). If the first option they generated was of random quality, it would falsify the RPD model. The results supported the RPD model — as the model predicts, the first option the skilled chess players consider is of high quality.
Also, the research supporting the RPD model has been replicated several times.
5. The evidence base is fairly thin. Some decision researchers have complained that there aren’t many studies conducted under controlled conditions. A government research sponsor once commented that she would never fund field studies. After someone reminded her that Darwin had relied on field studies, she replied, “And I wouldn’t have funded him, either.”
However, our field studies have shown their value for uncovering dynamics of decision making such as the role of experience. The RPD model describes how people use 10 or 20 years of experience. It is impractical to run controlled studies, employing unfamiliar tasks, and provide that amount of experience. In contrast, decision researchers relying on controlled laboratory conditions failed to grasp the importance of large amounts of experience for making effective decisions.
6. The model isn’t useful. I worried about its value when I first started publishing about it, but practitioners are clear that the RPD model has freed them from trying to apply rational choice techniques that don’t fit with complex and ambiguous situations. The RPD model guides a variety of training approaches for effective decision-making.
Recently, Gigerenzer (2019) stated that “… on its 20th anniversary, ‘Sources of Power’ continues to offer relevant directions and corrections of current research on decision-making.” (p. 479).
Gigerenzer, (2019). Expert intuition is not rational choice. American Journal of Psychology, 132, 75-480.
Kahneman, D. (2011). Thinking, fast and slow. Farrar, Straus and Giroux.
Klein, G. (1998/2017). Sources of power: How people make decisions. Cambridge, MA: MIT Press.
Klein, G. A., Calderwood, R., & Clinton-Cirocco, A. (1986). Rapid decision making on the fireground. Proceedings of the Human Factors and Ergonomics Society 30th Annual Meeting, 1, 576-580.
Klein, G., Wolf, S., Militello, L., & Zsambok, C. (1995). Characteristics of skilled option generation in chess. Organizational Behavior and Human Decision Processes, 62(1), 63-69.