This is really the heart of the issue, in 2020 at least (I hope people will go back to arguing about climatology soon!). In theory epidemiology does fall squarely within that definition. In practice if you read their papers you'll quickly discover most of them are entirely pseudo-scientific. I've read a lot of them this year. Here are some things that are absolutely standard in computational epidemiology (modelling) papers:

• Making predictions without writing a follow-up paper to study how closely observed data matched. Epidemiology routinely makes predictions that are wildly different to reality. This has no impact on the field at all. Science is meant to involve adjusting hypotheses if predictions don't match collected observational data, but in epidemiology step 1 is make a prediction with a model. Step 2, goto step 1.

• Making unfalsifiable predictions.

• Ignoring data that falsifies predictions.

• Altering predictions post-publication via public statements.

• Truncating graphs to hide data that would invalidate an argument.

• Using long since obsolete data, or data that aren't the best available when the paper is written. Invariably this makes the predicted epidemic worse, I've not yet seen anyone do this where it leads to better outcomes.

• Giving confidence intervals so wide they are equivalent to saying "we don't know what will happen" whilst not making that obvious.

• Incorporating model parameters that are simply made up, not based on any observational data.

• Engaging in circular reasoning.

• Ignoring basic biology. One of the most surprising things about epidemiological models is that they basically don't utilise any biology at all and epidemiologists will happily violate basic biological or historical common sense if it helps them achieve their goals. For example they calculate R0 using case curves. R0 is meant to be an intrinsic property of a virus representing how infectious it is, but one model I saw calculated a totally different value for COVID for each outbreak/country it was applied to. Clearly their values for R0 couldn't be correct, as it's the same virus everywhere so the values should be at least a little bit consistent. They were orders of magnitude different! This didn't bother anyone in the modelling team or anyone who peer reviewed their work at all. Another paper assumed Rt (case curve rate of change) could only change if government interventions forced it to, which flatly contradicts tens of thousands of years of experience in which epidemics rise and fall naturally, without lockdowns or masks of any kind. No peer reviewer quibbled with this, one even said it was a reasonable assumption.

I could go on all day. Maybe one day I'll write this up somewhere outside HN. But seriously, epidemiology is by far the most problematic field of "science" I've ever seen, and I read plenty of research papers from other fields. I don't consider what they're doing to be even slightly scientific: way too many papers are flat out intellectually fraudulent.