Automatic Decision Making

autoimagesI, in a relatively clueless sense, have embraced the idea that automatic processes, intuitive, System 1, etc. make a lot of our decisions, especially that automatic processes are for output and deliberative processes are for input. (Intuition in J/DM)  The paper titled: “Deliberation versus automaticity in decision making: Which presentation format features facilitate automatic decision making?” was written by Anke Söllner, Arndt Bröder, and Benjamin E. Hilbig and appeared in the May, 2013, issue of Judgment and Decision Making, tries to rain on my parade, at least a little.

The authors adopt the view of the multiple strategy approach, that people are equipped with a repertoire of different decision strategies, but when you do this the strategy selection problem arises:  How does the decision maker determine which strategy to choose?  Sollner et al say that Marewski and Schooler observed that the strategy selection problem is frequently resolved by the fact that sometimes only one specific strategy (or at most a small set of strategies) is applicable or afforded. In their  cognitive niche framework Marewski and Schooler assume that strategy selection follows a cost-benefit-tradeoff only if more than one strategy can be applied. Hence, according to this framework, the task environment already constrains the set of applicable strategies, thus facilitating cost-benefit selection between the remaining ones.

Glöckner and Betsch avoid the strategy selection problem by assuming only one uniform mechanism, parallel constraint satisfaction.  Sollner et al however, deem it sensible to treat the PCS model as if it belongs to a repertoire of decision strategies for several reasons: (1) Proponents of the PCS model repeatedly treated it themselves as if it was one of several applicable strategies by contrasting it with the different decision strategies instead of the multiple strategy approach as a whole. (2) Some evidence from these investigations suggests that the PCS model cannot successfully account for decision making under certain constraints (i.e., inferences from memory, forced sequential information search). (3) Despite PCS’s notion of parallel information integration, evidence has been found that people sometimes actually employ TTB(Take the Best) in a serial manner—a finding, PCS cannot easily account for. Finally (4), even a unifying model can account for differences only by assuming different parameter values. How these are determined is a structurally similar problem to strategy selection in a multiple strategies framework.

In the course of monitoring how different task features influence people’s strategy selection (within the multiple strategy approach), the presentation format has been addressed by several authors. These findings suggest that not the presentation format per se, but the accessibility of information, as determined by the presentation format, influences which strategies decision makers employ. In their investigations, the proponents of the PCS network model have predominantly employed the matrix-like presentation format of the open information board which demands particularly little information search. Here, a remarkable dominance of PCS-consistent behavior has repeatedly been shown. In turn, one might question whether presenting information in an open, matrix-like format is actually a necessity for PCS. Indeed, this is plausible given that one crucial pre-condition for “automatic” decision making (in line with PCS) was mentioned by Glöckner and Betsch: sequential information search seems to impede the reliance on PCS.

To test these ideas, the authors conducted three experiments using different information presentation and determining whether subjects used PCS or one of the sequential decision strategies.

Employing the City-Size task for probabilistic inferences, the authors developed an alternative presentation format (based on the idea of a map) to test the assumption that decision makers’ ability to rely on PCS-like processes is bounded by information accessibility and thus hampered once the “wrong” presentation format is used.
Across all three experiments, a robust presentation format effect between the matrix with high accessibility of information and the map with reduced information accessibility could be found: PCS-compatible behavior was much more prevalent in the former than in the latter, whereas participants used one of the sequential decision strategies (WADD, TTB, or EQW) more frequently in the map than in the matrix.  Only when all pieces of information are instantly and simultaneously available, PCS-consistent behavior is predominantly observed.

The important role of information accessibility to PCS consistent behavior was only recently discussed for inferences from memory. Comparing their research results
with the findings of others, Glöckner and Hodges concluded that the accessibility of information might constitute a relevant variable that influences the process of decision making. When all applicable pieces of information are quickly available without high memory costs, PCS-consistent behavior can be observed, whereas it is only rarely found when retrieval imposes high memory costs and information accessibility is therefore reduced. Our results show that this reasoning can be transferred to inferences from givens as well. The current results extend such conclusions in showing that—even in presentation formats that display all pieces of information simultaneously—a moderate reduction in information accessibility also reduces the prevalence of PCS-consistent behavior.

But why does the ability to integrate information in a PCS-consistent manner seem to crucially depend on a high accessibility of information? Glöckner and colleagues emphasized that the PCS model integrates automatic, perception-like processes. They explicitly draw a parallel between their PCS network model and Gestalt psychology’s basic idea of automatic consistency maximization.

PCS-consistent information processing relies on this immediate constitution of a mental network and is thus hampered when information needs to be restructured and recoded before it can be integrated. If the quick, automatic default for decision making (PCS) cannot be applied, the decision maker has to resume to sequential decision making strategies instead.  According to the authors, this reasoning is well in line with Marewski and Schooler’s cognitive niche framework: For different environments, different processes are applicable. From a cost-benefit-view, “automatic” decision making should prevail whenever it is applicable. If, however, the constitution of the proposed mental network is impaired, the default strategy is no longer applicable and people have to select a different decision strategy from the set of applicable options. The findings of this paper emphasize the importance of considering both information acquisition processes on the one hand and processes of information integration on the other hand as two interdependent but nonetheless separate parts of the whole decision making process.



2 thoughts on “Automatic Decision Making

  1. Pingback: Strategy Selection - Single or Multiple? - Judgment and Decision Making

  2. Pingback: Cognitive Continuum - Judgment and Decision Making

Comments are closed.