Ask HN: What parts of software engineering can be quantified? | Hacker News Reader

Ask HN: What parts of software engineering can be quantified?

2 pointsbordercases2 comments

I've seen obvious instances of quantification of software engineering – obvious in hindsight, anyhow. Capacity planning and network performance get strong mathematical treatments in textbooks such as "Computer Architecture: A Quantitative Approach".

But I've also seen less-than-obvious quantifications and mathematical models. The second chapter of "Hello Startup!" represents a model of Agile vs Waterfall method in terms of total cost-under-curve for learning, complete with graphs.

I've also seen the process of debugging code be represented in Bayes Theorem, but that was only one problem in a textbook.

And of course anything by Knuth or Sedgewick tends to carry this flavour with their takes on data structures. But that's moving a bit far from software creation in the mobile-app and web-app sense.

What are examples of practical mathematical modelling and quantification in software-engineering?

It depends on the degree of quantification and your precise definition of software engineering, but IMHO almost all of software engineering research is minimally as quantitative as the social sciences i.e. psychological or sociological studies and analyses. In other areas, as you also mention, SE is almost purely quantitative in nature.

see: https://jserd.springeropen.com/, https://research.google.com/pubs/SoftwareEngineering.html, and https://arxiv.org/list/cs.SE/recent for some recent SE research topics.

Software Engineering encompasses, broadly speaking, the following topics:

Software Design => Implementation => Testing & QA => Automation => Project Management

Implementation, Testing/QA, and Automation are largely very quantitative, you can create models and algorithms and analyze them and measure most things even easier than in the physical sciences. On the other hand, due to the nature of Software Design and Project Management you are largely left to tools of the social sciences i.e. surveys, difficult to control experiments, observations, etc.

What parts of thought can be quantified? Bad questions get bad answers.

Reply on news.ycombinator.com