EPREVAEngineer Paper
OpinionComputing & AI for engineeringOpinion — the author's view

Engineering papers should run, not just be read

Most engineering knowledge is published as static PDFs that can't be checked, changed or built on. It's time to publish the model, not just a picture of it.

A laptop showing an interactive plot next to printed research papers on a desk
AI-generated image

A typical engineering paper contains a model — a set of equations with parameters — and a few figures showing what the model does for a handful of parameter values. The reader gets the pictures. The model itself, the thing that took months to build, is left as an exercise.

That made sense when papers were printed. It makes much less sense now. Data from PLOS ONE studies show how fast the supporting material around papers decays: about one in five science articles suffers from reference rot, and supplementary data hosted inside PDFs is far more likely to disappear than data with a DOI.

What "running" means

It does not mean every paper must ship a software package. It means three modest things:

  1. The governing equations are live. If a paper says ΔT = R_T·Q, the reader can change R_T and Q and see what happens — like the playground below.
  2. The units are checked. Every rearrangement of a published equation can be checked for dimensional consistency by computer. We do this for every formula on Engineer Paper.
  3. The data and code have a permanent address. A DOI from Zenodo is free. A GitHub link is not an archive.
Playground · Temperature rise from thermal resistance
solve for
Temperature rise
10 K
Plot temperature rise against
1,00000.01200 K/W (log)

Why this is not just nice-to-have

Interactive papers find errors. A slider that produces a negative temperature, or a curve that bends the wrong way, catches a sign error faster than any reviewer. Replication gets easier when the model is in front of you.

What we are doing about it

Engineer Paper's Submission Studio lets authors turn equations into interactive models with one line, records which parts of a work are reproducible, and keeps every version. It is a start. We would rather publish fewer papers that run than many that only sit there.

Use it

Sources

  1. Scholarly Context Not Found: one in five articles suffers from reference rot — PLOS ONE
  2. Availability of supplementary data and links in scientific articles (2024) — PLOS ONE

Discussion

Technical questions and corrections are welcome. Comments are signed and checked before they appear.

No comments yet.

Sign in to add a typed review comment. Reviews are signed with your name.