Rules for anyone writing lessons, stories, timeline entries, or Bug Hunt challenges.
When we describe what a researcher did, we use the words they used, quoted and cited.
Michie (1963) gave MENACE a box for each of “the 287 essentially distinct positions which the opening player can encounter” (source). Most modern rebuilds use 304 boxes.
McCulloch and Pitts called their units “neurons”; today we’d call them threshold logic units.
Key terms, in their words (each one checked against the original paper):
| They said | Who, where | Today we say |
|---|---|---|
| “neurons” with “all-or-none” activity, treated “by means of propositional logic” | McCulloch & Pitts 1943 (DOI) | threshold logic units |
| “perceptron”: “a probabilistic model for information storage and organization in the brain” | Rosenblatt 1958 (DOI) | single-layer linear classifier |
| “Two machine-learning procedures have been investigated…” | Samuel 1959 (DOI) | machine learning (one of the term’s earliest uses) |
| “(soft-)search”, “align”, “alignment model”; “attention” appears only twice | Bahdanau, Cho & Bengio 2014 (arXiv) | attention |
| “the 287 essentially distinct positions” | Michie 1963 (DOI) | 304 boxes in most modern MENACE rebuilds |
| “Shortcuts are decision rules that perform well on standard benchmarks but fail to transfer…” | Geirhos et al. 2020 (DOI) | shortcut learning |
Why: paraphrasing is how history gets bent. A real result gets attached to the wrong machine, the wrong decade, or a bigger claim than the authors made. Quoting the source keeps us honest. It also shows students how to check claims, which is the point of the Bug Hunt.
Studies that classify people (by sex, race, faces, and so on) are history we can discuss, quoting the authors’ own terms. They are never hands-on student activities. We treat them as conversations about bias and fairness.