[marginalia_nu]

The ones who adopt a scientific approach are by far the worst. Design is ultimately all about how things ought to be, an act of judgement, meanwhile science is wholly unsuitable for such questions, since it only tells us what is, which following Hume, cannot on its own lead to conclusions about what ought to be.

You get a sort of garbage-in-garbage-out effect if you apply science to a field like design, where it only serves to amplify your own convictions, as what is being fed into the scientific process as unquestioned assumptions inevitably fall out of it as conclusions.

At best you get KPI driven design, which is a vehicle for enshittification, not for building great design.

[hotnfresh]

I'm not following. Science gives us good guidance on what works well or will let us achieve our goals, all the time. It's basically the whole point of doing it at all.

I took the poster as meaning UX that considers the results of, and perhaps even performs, actual user testing & observation, to decide what works and what doesn't. Like operating system vendors used to. I'll grant that "scientific" UX that's just incompetent (99% of the time) application of "telemetry" and A/B testing is awful. But that—and the other bad kind that's just trend-following, personal preference, and whatever will get the best reaction in a design presentation meeting full of non-experts—aren't what I understood as being advocated.

The good kind performs & pays attention to science.

[marginalia_nu]

What science doesn't give good guidance is how to select those goals in a vacuum. The goals at best end up being a version of someone's personal opinion, since there that's the only form of opinion we have access to.

Any opinions you get out of the scientific method were put in there by the person designing the experiment.

[bigDinosaur]

The core of science is iteration based on experiment and hypothesis generation. The scientific approach to UX is simply a hypothesis that a UX change is superior, and an experiment with a measure as to whether or not this is true. Yes, of course you can optimise scientifically for user-antagonistic KPI's (no need to invoke Hume, his point is true and also useless) but the alternative to not adopting a scientific approach is literally just opinion which on the whole is far worse (would you want your airline cockpit's UX to be designed based on a designer's opinion, or real scientific approaches to how people process information and intuit controls?). Of course one would take a good designer's opinion over some MBA-ridden process but a good designer is likely intuiting what would be validated by a more scientific approach anyway.

[itishappy]

Can you relate that to specifics from the article? To my (admittedly non-designer) eyes it appears to be a great example of how science can be used to improve design, and I happen to agree with the findings presented, so I'm curious where you see this breaking down.

[marginalia_nu]

The tacit assumption being made in the article is that good design shouldn't frustrate the users.

I don't disagree with this, but it's none the less an assumption that went into the study, and likewise a conclusion that fell out of it.

[bigDinosaur]

Nobody here is disputing we have values that are not scientific.

[jfim]

> The ones who adopt a scientific approach are by far the worst

I strongly disagree.

Design without considering all of the HCI research that has been done is what you call "garbage-in-garbage-out." We already know how humans perceive information, what makes things salient or invisible, and so on, yet the current design trends completely disregard that with flat UIs and trendy designs that have poor usability.

> At best you get KPI driven design, which is a vehicle for enshittification, not for building great design.

No, you just get trendy design, not usable design.