Formal reasoning, intuition, and embodied cognition in decision making

Thinking/reasoning of both humans and machines could be arranged along the formality gradient of semantics. This picture is adapted from Anatoly Levenchuk’s book Education for Educated (in Russian):

The formality of thinking. If you have never seen the word “schemoid” before, consider this very picture: it looks like a schema, but is not as formal as a circuit schema of an electronic device, for instance.

Amazon’s six-pagers are tasked, among other things, to make decision-making more formal. Making decisions after an oral discussion is not completely intuitive (human language itself already has some structure and, therefore, is not completely intuitive), but is closer to the intuitive end of the spectrum than making decisions by wetting a written document (six-pager, RFC, ADR, etc.). In the post “Alphas of project decisions”, I suggested a way to formalise decision making even further than in Amazon’s six-pagers or Google’s Design Doc templates.

More intuitive thinking is faster and energetically cheaper but is prone to more errors than more formal reasoning. Therefore, the formality of the decision-making process should be adjusted to the cost of error for this-or-that decision.

However, even when one follows a more formal process for making a particular decision, it’s important not to forget or disregard intuitive thinking completely.

Idea for a startup: AI to predict the success of a movie by the script

Machines excel at both ends of the spectrum: intuitive thinking (distinguishing cats from dogs) and purely formal reasoning (source code execution, calculation, theorem provers), but are weaker than humans in the middle range of formality.

This means that intuitive evaluation of decisions could be delegated to AI relatively soon.

An AI model (which would include, or would be based upon some language models) could be fed with scripts of all movies that have been made and their eventual success (e. g., the Metacritic rating, or the box office to budget ratio). Then the model could predict the future success of a movie by the script.

With a little longer stretch of the imagination, this idea could be applied in business, too. An AI model could be fed with press releases of many products and services that have been launched in the past years by various companies, together with the metric of their eventual success (e. g., sales). Then the model could predict the future success of a product by the press release which is written before the product is built, at the product discovery stage, according to Amazon’s PR FAQ process.

Embodied cognition in decision-making

Related to intuitive thinking, but not equivalent to it is the concept of embodied cognition. Embodied cognition is sometimes important in business, too! Here’s an interesting example from Colin Bryar and Bill Carr’s book Working Backwards about Amazon:

We also constantly evaluated the “form factor” of the Kindle—the size, shape, and ease of use—during the iterative design review process. The first prototypes were nothing more than Styrofoam cutouts with mocked-up screens and keyboards. As the form took shape, we evaluated models made of plastic that were weighted so the shape and feel would be as close as possible to the real thing. At every review, Jeff would spend several minutes holding each prototype in one hand, then the other, then in both. When he rejected a prototype it was typically not because the design wasn’t sleek or hip enough but rather because something about it would “get in the way” of the customer’s reading.