JOURNAL ARTICLE · EDITORIAL BRIEF
Quality versus quantity: including imperfect training data improves prediction of user satisfaction in software projects
Łukasz Radliński
Knowledge and Information Systems · 26 October 2026
The finding or development
Models retaining more lower-rated project records generally predicted the study’s satisfaction score better than models trained on smaller, cleaner subsets. Filtering data can discard useful variation as well as errors.
What supports it
Peer-reviewed computational comparison: 134 projects, 60 data variants and 12 predictive models, with repeated nested cross-validation.
Reviewed: Publisher abstract, dataset preparation, evaluation methods, findings and validity discussion; code and supplementary results not audited.
Why it matters for design
A methods reading for testing data-exclusion rules rather than assuming stricter filtering always improves prediction. It does not justify accepting weak evidence in a literature review.
Read with this in mind
One small, voluntarily submitted historical dataset; satisfaction was an unweighted sum of eight attributes. Random rather than temporal splits and target-informed preprocessing warrant scrutiny before applying the rankings elsewhere. Prediction accuracy does not identify design improvements or replace user research.
Original source
Opens the publisher outside this reading room. If nothing opens, copy the source address below.
Reference tools · Paperpile & more
Using Paperpile? Open the original source above, then use the Paperpile extension to save it to your library. Its inline buttons appear on supported websites.
Paperpile’s browser extension guide (opens an external website)Or download the original reference and import the file into Paperpile, Zotero or EndNote. Check the details against the source before citing.
Citation details are requested from Crossref when you download. The file contains a reference, not the full paper.
DOI: 10.1007/s10115-026-02898-9
This is the complete UX Observatory brief, not the publisher’s full article. Follow the original source to read the work.