(a first part summary of essays contained within Gigerenzer and Selten's [eds.] Bounded Rationality: The Adaptive Toolbox)
What will I make for dinner? A relatively simple question that plagues me most nights of the week. However, what is a relatively simple task, albeit one that can be arduous for me, unpacks a can of worms in basic research: through what mental process do I decide what to make?
Admittedly this is a silly question, but other questions used to motivate this issue are roughly as silly. The point is merely to illuminate the difference between decision-making paradigms. Classically, or as far as my knowledge extends to, we would typically think of solving this problem through maximizing our utility, searching for the best possible answer (the global optima). In the context of our dinner problem, the steps to solving our problem would look like:
- mentally assemble all of the possible recipes
- determine which recipes had all of the possible recipes
- determine which recipes had all available ingredients
- account for the spoilage dates of all the ingredients
- identify the meal that would provide the maximum utility (one that uses the most ingredients that are about to spoil the soonest while being feasible to make)
The reality that this decision methodology ignores are the constraints in the real world. When I start to consider what to cook, the entire process has typically started because I'm hungry. This imposes a very real time constraint because if too much time passes I become what is known as "hangry" (both "hungry" and "angry" at the same time). In the process of maximizing my utility I generally ignore this time constraint and instead suffer in the search for the optimal answer. The situation considerably worsens for me if I am a poor cook and have few or no recipes memorized, since this introduces researching online first to identify possible recipes. There is also the additional kink of, how do I assess when I have spent enough time finding recipes so that the optimal recipe would be in the list of recipes I now possess? I could continue to look further and make comparisons, but this could rapidly lead to a situation where I am continually conducting research to determine if my prior research is sufficient. This leads to decision paralysis and the possibility of infinite regress, i.e. where I just keep going further down the rabbit hole of preparing to make my decision without making any progress towards resolving it.
Crap. At this rate I might not start eating dinner until tomorrow or next week!
This doesn't really seem to reflect my typical reality though. There is a fairly serious time crunch, wherein I must decide before hanger sets in.
In the early 1950s, Herbert Simon set out an alternative theory of satisficing (stemming form the concepts of satisfy and suffice). Using this heuristic we actually go about decision-making in a different manner. Within this paradigm I instead set some minimum value of acceptability and instead evaluate each possible dish as I think of it. As I evaluate some dish I calculate its score based on the number of ingredients it uses that are about to spoil. As soon as I think of a dish that has a score greater than my set point I start cooking. I don't worry if I have the most optimal dish, I instead just start cooking.
This is where the bounds on rationality come from, I use a heuristic to achieve a locally optimal solution since it is more than likely good enough and fulfills normal criteria (like a lack of unlimited time). A fully rational agent (in the economic sense) would seek the optimal choice in the situation. An irrational agent would choose at random. A last class, which some like to equate to bounded rationality but differs, is a ration agent under constraints. We could say that in our dinner problem I am aware of the available time constraint and use that as a factor in my utility. However, in this situation I am still seeking the global maximum utility within the constraints of my problem, which differs from my heuristic of take the first answer that satisfices my concerns.
And at this point I'll conclude my first summary of the works. From here we start exploring the role that the environment plays in decision-making, and our ability to exploit it when it doesn't change rapidly with heuristics, but that seems like a good second topic.