Cognitive profile

A structured profile of one real organism's cognition, answering nine fixed sections, and closing on what that organism lets you see about people that you cannot see from inside.

Due: 21 May 2027 · Weight: 40%

Choose one real organism and profile its cognition against the nine sections below. Each section asks for a different kind of claim, and that is the point of having them — a paper gives you results, and the profile makes you decide whether each result is a specification, a limitation, or a warning.

The sections are fixed. The presentation is yours. A document, a poster, an illustrated profile, a web page — whatever suits the organism, as long as every section is answered and a reader can find them.

The sections#

1. The organism, and why this one. Short.

2. Sensory channels. What it detects, with range and resolution where known — and what it cannot detect at all.

3. What it computes, and what it stores. Not what the animal does. What representation it holds.

4. What it does better than us. Every system here buys something; say what.

5. What it cannot represent. Not “is poor at”. Has no representation for.

6. Demonstrated failures. Real experimental results where the system fails.

7. Which of those failures have no fix, and why.

8. What it cannot detect about itself.

9. Comparison with people. Below.

10. Optional, and speculative. Either one change to the organism and the consequences you can derive from it, or one system the model was never derived from — an artificial one, or a mind that does not exist — with what it could not represent and could not detect about itself, derived before you look it up. Label it [S].

The tenth section is the only place on this course where you may reason past the evidence on purpose. It is marked on whether the derivation follows from the model rather than on whether the guess is right, so a confident answer that does not follow scores worse than a tentative one that does.

Section 3 is the hard one#

Students want to write behaviour here. Write representation.

Cataglyphis stores a vector, not a route — a bearing and a distance to the nest, continuously updated. Which is why the ant always knows where home is and never knows how it got where it is. That sentence is a claim about what is held in the animal, and it is the sort of thing this section is for. “It navigates accurately” is not.

Section 7 is where weak profiles collapse#

Everything looks fixable to someone who has not understood the mechanism. The distinction between fails under these conditions and cannot, in principle is the distinction the whole course depends on, and you need at least one of the second kind with the reason attached.

The ant’s integrator has no independent reference against which to check its own output. No amount of better engineering inside that design removes the problem, because the checking would have to come from outside the thing being checked.

Section 9 — the comparison#

The obvious version of this section is bad. “Humans are better at X, the organism at Y” is a ranking, which is the thinking this course rejects, and it requires nothing of you.

What this section asks for is one structural parallel — a limit you documented in the organism that has a human counterpart of the same shape, not the same magnitude.

The ant cannot detect its accumulated navigational error, because the system doing the checking is the system being checked. A lost ant behaves exactly like an ant that is not lost. You cannot detect your own miscalibration, for the same reason. Different substrate, identical structure.

So the question to answer is: what does this organism let you see about human cognition that you could not see from inside it? That is the course’s method in one section. You can see the ant’s bound from outside. You cannot see yours from inside. The organism is how you find out what shape yours might be.

Do not rank the two systems. State what each trades.

What fails#

• Behaviour written where representation was asked for.
• Section 7 with everything marked fixable.
• A comparison that ranks, or that finds a difference rather than a parallel. Differences are easy and say nothing.
• An organism chosen because it is charismatic rather than because there is literature to support the sections. Octopuses are over-chosen and under-evidenced every year.

Choosing well#

Difficulty varies more than you would expect. The desert ant is tractable: one system, decisive experiments, clean limits. The mantis shrimp is harder because its specifications contradict the obvious reading. A honeybee swarm is hardest, because the unit is a colony and the reader keeps forgetting.

Bring two options to Rosalind before the end of week 9. She will usually tell you which one has the literature to fill section 7, which is the section that decides your mark.

The rules that apply to this task#

Extensions, late penalties, academic integrity and what you may use a language model for are on the policies page, which governs all four assessments. Nothing on this page overrides it.