Reader styles
The standard is the best current guess for a reader nobody knows. A reader style says what moves once the reader is known, for whom, and on what evidence.
The best choice does move: what helps a novice slows an expert down (expertise reversal: an aid that helps beginners hurts readers who no longer need it; Kalyuga et al. 2003), and a model reads differently from a person.
Contents
| File | The reader |
|---|---|
| template.md | blank style to copy for a known reader |
| novice.md | new to the domain; needs every term, worked steps, explicit links between ideas |
| expert.md | knows the domain; reads for the new part, by search, often out of order |
| visual-relational.md | prefers relations as graphs, sets as regions, quantities as geometry |
| llm-reader.md | a language model reading the text as context |
How a style overrides the standard
- A style only lists overrides: principles it strengthens, relaxes or replaces. Everything else in the standard holds.
- Each override names its principle, says what changes, and cites why.
- A style may also change a check's thresholds; it does so in the repository's
checks/config.toml, never by editing the check. - A style for one named person stays with that person. The ones here are archetypes.
Example
A repository written for experts sets trigger = 120 in checks/config.toml and keeps worked examples to the first use of each method, as expert.md says. Run python3 checks/run.py . and the logical-lines questions now start at 120 statements.
What a style cannot do
The meshing hypothesis, matching the presentation to a preferred learning style, does not improve learning in the experiments that tested it properly (Pashler et al. 2008).
So a style states preferences and known expertise, and promises nothing from matching alone. See evidence/learning-science.