Ill-Structured Domains Aren't Necessarily Wicked

This is Part 4 in the Learning in Ill-Structured Domains Series. Read the previous part here.

This is a companion discussion topic for the original entry at
1 Like

Chris Wong commented on something I wrote and cited this piece.

While writing out my response, I realized perhaps I should address my questions here instead. So here it goes

And so here’s my point: ill-structuredness is a separate property from the kindness or wickedness of a domain. It is totally possible to operate in a well-structured domain but within a wicked learning environment (where feedback is — for some godforsaken reason — withheld from you), or in an ill-structured domain but within a kinder learning environment (where it is relatively easy to track the outcome of decisions on highly variable concepts over long periods of time).

I observed the following distinctions:

  • property1: kindness with the possible values: kind / wicked
  • property2: structured-ness with the possible values: well structured / ill structured
  • class1: domains like investing, business, etc these tend to be abstract
  • class2: environments which is a smaller concrete “space” than domains. these tend to be more concrete than domains.

Cedric seems to say property1 and property2 are different. But I am not sure how they are different. Or more importantly, how they are related.

The confusion is especially acute here.

Most importantly, though, Hogarth wrote that you may influence the validity of your intuitions — and the way that you do so is by picking the sorts of learning environments that you develop your expertise in.

In other words, kind learning structures and wicked learning structures are orthogonal to the kindness or wickedness of an operating domain.

First, Cedric cited Hogarth (I also haven’t read)
Then, he took Hogarth’s learning environments and turn into learning structures.

I am unsure if Cedric means

M1: “property1:kindess learning environments are orthogonal to the property1:kindess of a domain”
M2: “property2:structuredness learning environments are orthogonal to the property1:kindess of a domain”

Also there’s another possibility I have considered when Cedric chose the words kind learning structures instead of the original Hogarth’s words “kind learning environments”

M3: at the environments level, property1:kindness and property2:structuredness are one and the same. but at the domains level, they are different.

M4: at the environments level, both property1 and property2 are orthogoanl to property1 at the domains level. property1 and 2 are different regardless at which class level (environments vs domains)

So which of M1, M2, M3, M4 does Cedric mean?

Hogarth uses the words learning structure and learning environment interchangeably.

The two things (ill/well structuredness and kind/wicked) are unrelated.

  • Ill-structured domain: there exists concepts, but instantiations of those concepts are highly varied. Novelty is the norm.
  • Well-structured domain: there exists concepts, and instantiations of the concepts are identical.
  • Kind learning environment or structure: Is feedback clear, is learning easy? (There’s a longer list of questions but that’s listed in the blog post).
  • Wicked learning environment: Is feedback complicated and learning hard?

Medicine is an ill structured domain. But there can be kind and wicked learning environments in it:

  • Kind: surgeons have a kind learning environment. If they make a mistake, the patient dies, either on the operating table or not long after, with the surgeon being present for the outcome.
  • Wicked: emergency room doctors triage and hand-off to other doctors. They don’t typically follow up and therefore don’t have feedback for whether their decisions were the right ones.

Investing is an ill structured domain. But there can be kind and wicked learning environments in it:

  • Traders have fast feedback loops and the scaffolding of decades of judgment and decision making research to help them; such research is usually part of the training programs and simulations they are put through.
  • Long only investors can take decades before they figure out if they actually have skill, or if their returns are due to luck.

You can carve out kinder learning structures inside an ill-structured domain. For instance, certain long-only hedge funds do case studies that compress investing decisions that would normally take place over the course of a decade into the space of an hour.

Similarly, high school math is a well structured domain but I can trivially create a wicked learning environment for you: I’ll just hand you a scoring rubric that scores your answers randomly for every test and I send it to you via post two months after you complete a mock exam.

1 Like