A transcript of Episode 228 of UX Podcast. James Royal-Lawson and Per Axbom discuss The Aesthetic-Accessibility Paradox and Three Principles for Designing Machine Learning-Powered Products.
This transcript has been machine generated and checked by a human.
Transcript
Per Axbom
UX podcast is funded by James and myself together with any contributions we can get from you, our listeners. You can contribute any amount you like, whenever you like, by visiting UXPodcast.com/support.
Computer voice
UX Podcast Episode 228.
[Music]
Per Axbom
Hello, I’m Per Axbom,
James Royal-Lawson
and I’m James Royal-Lawson,
Per Axbom
and this is UX Podcast. We’re in Stockholm, Sweden and you’re listening in all 50 states of the USA from South Carolina to North Dakota.
James Royal-Lawson
And today is time for a link show. We have- yes- two articles for you, that Per and I have stumbled upon, staggered over, and read recently and with-
Per Axbom
During our digital travels.
James Royal-Lawson
Ah, you normally do say that, don’t you? During our digital travels.
Per Axbom
We always have to say that.
James Royal-Lawson
… digital travels during the festive period. So.
Per Axbom
Right.
James Royal-Lawson
The, the two articles are The Aesthetic Accessibility Paradox by Anthony Zang, on UX Movement.
Per Axbom
And the second article is Three Principles for Designing Machine Learning Powered Products by Mat Budelman and Mark Kizelshteen or Kizelshteyn.
James Royal-Lawson
They work at Spotify Design.
Per Axbom
But let’s start off with the, the aesthetic accessibility paradox because I think we, we have a lot of feelings about this, and I think one challenge here is not to get too involved and emotional in some of the ways that, he, the author expresses how he looks upon accessibility because there you can read the article and you can look at the comments. The comments section is quite revealing as well. But let’s start off with what this is about.
So there’s essentially, according to author, the author, a conflict between making something visually appealing and making it accessible for just everyone. And he, he goes on to describe this using a vision as the example. So, in the first paragraph, he, he claims that the majority of users usually have normal vision is his, is his, sentence there, while the minority have some form of visual impairment. And so the struggle here is that if you actually don’t have a visual impairment, he says, then you want it visually appealing, but if you want it accessible, it won’t be visually appealing for the majority of users, which will make it a worse experience for those people-
James Royal-Lawson
– a lesser design.
Per Axbom
That is the essential argument. Exactly.
James Royal-Lawson
But- But, first off, we’ve got, we’ve got to query the, the, majority and normal vision part of it.
Per Axbom
Yeah, let’s start off by questioning that assertion in the first paragraph: the majority of users usually have normal vision.
Per Axbom
Now, I think that’s-
Per Axbom
I know you-
James Royal-Lawson
I think that’s quite true.
Per Axbom
Yeah…
James Royal-Lawson
But, I’ll say that with a star next to it, because when we’re talking about the minority that do have a visual impairment, the World Health Organisation, they’ve recently claimed, or said, that over 2.2 billion people worldwide have a visual impairment, treated or untreated.
Per Axbom
And in that case, it actually is the minority then. It’s 2.2 billion, but it’s a minority.
James Royal-Lawson
– minority. It’s not a minority to sniff at because 2.2 billion people is a lot.
Per Axbom
Yeah. What I did though, was actually found a study from 2005 that tells us more- it’s from Jobson Research- tells us more about the state of vision impairment in the USA. And according to this study, over 61 percent of the population, or 177 million people in the US need some sort of vision correction. And that, to me, sounds like there’s actually a majority who do not have normal vision. We could also start talking about- we can’t get into it, specifically everything- but why would you call it normal vision? Why is it normal? If it’s not necessarily the majority as it is found by this study.
James Royal-Lawson
You know, there’s
James Royal-Lawson
Yeah, we, we-
Per Axbom
And, but-
James Royal-Lawson
There’s problems, there’s problems with that phrase “normal,” because it’s normal from a certain perspective…
Per Axbom
Which implies that something else would be abnormal. But there’s another interesting figure from this study: 12.2 million adults require some sort of vision correction, but don’t use any, so you, you, might require glasses, but you don’t use it, to 12.2 million adults in the US alone.
James Royal-Lawson
So, but if we’re moving on from-
Per Axbom
We’re talking about a lot of people.
James Royal-Lawson
We are talking about a lot of people, but from moving on from the, the, stats of visual impairment, the- one of the quotes from the article is: in general, the more accessible an interface is, the less aesthetic appeal it will have. Now, that, I just don’t know where to begin with that sentence, because, I mean, how do you, how do you, how do you claim that? I mean, how can you claim that if, if, if an interface is accessible, they made to be fully inclusive, that it’s at ugly?
Per Axbom
Right.
James Royal-Lawson
But look the crux of this, I guess, is you’ve got one thing, aesthetics, which is, in my opinion, subjective, beauties in the eye, in the eye of the beholder, it’s, it’s, it’s entirely subjective, is who, is what you think. And then we’ve got accessibility, which to a large extent, is something objective.
Per Axbom
Right.
James Royal-Lawson
You can-
Per Axbom
You can measure if somebody has access to the solution or not.
James Royal-Lawson
Yeah. And on many, many access- aspects- of accessibility, you can literally say yes or no. It is, filled, it can’t do that you can’t do that you can’t access that you can’t do that. You can’t do that. And even with colours, I know this is, can be, a bit of a contentious topic as well, but you can measure contrast, you can measure readability, you can develop ways of actually measuring whether something is likely to be accessible from a, from a, colour perspective. When it comes from-
Per Axbom
So, what we’re saying, really, is that two, two, versions that are very aesthetically pleasing for some people, because we know it’s subjective. Both one of those could be inaccessible one of it could be accessible, but you can’t tell by just looking at it, whether it is.
James Royal-Lawson
No.
Per Axbom
You actually have to do the due diligence of finding it out.
James Royal-Lawson
Yeah, I don’t think so. I mean, how do you measure that aesthetic side of things? I mean to decide whether your design is highly aesthetic would require a large amount of research to basically, find, interview people and to specifically question, lay the question to them, about beauty- define beauty. Philosophers have been trying to define beauty for thousands of years and no one has completely managed to do it. So how the, how the, devil do you gonna manage to go out there and, and, and interview someone to get an actual result on a checkbox of whether your site’s of design is actually beautiful or not, just, we can’t do that.
As a bit of fun, I actually asked my 11 year old. I showed him this- in the beginning of the article, there’s a couple of pictures of a form on a website. Simple sign-up form, and they’re compared side by side. And one of them is about, is actually marked as highly accessible. to speed you just look more contrast heavy. And then the other one is marked as highly aesthetic, and it looks a bit less contrast absolutely compared to the one on the left. So I asked my 11 year old, “Which one do you like best?” shot in them. He wouldn’t understand the, this, he wouldn’t really bother reading the marking underneath. And, he goes, “that one” and points to the highly accessible one. So I said, “why that one?” And he goes, “because you can see it better.” The other one is just great.
James Royal-Lawson
Now, what, what I think was fascinating about that, is, I mean, he, he did a an aesthetic evaluation of these two features spontaneously without prompting, just kind of like, I threw them at him and said, and he chose the highly accessible one. Now Anthony, the article goes on, he shows a third one just, like, just below the fold on that one. Which is, and reveals that the, the two first pictures, one of them was AA compliant according to VCAG- WACG- and the other one was non-compliant. So the AAA versus non-compliant, and then he goes on to show AA compliant example, and says this one is the, is the balance.
Per Axbom
Right.
Per Axbom
And I think, I think that’s what, what upsets me most about this article is that it’s misleading in a lot of ways. It mis- misleads in the sense that aesthetics is subjective. What is more accessible could be more aesthetically pleasing for a lot, lot of people. There’s no evidence presented that it would, that would not be the case. He asserts that when it’s aesthetically pleasing, it’s harsh for people who have accessibility needs. And he, he says, contrarian that, if it, if it’s highly accessible, that it’s very harsh for people who have normal vision. I’m, I’m-
James Royal-Lawson
Yes!
Per Axbom
-doing finger towards him.
James Royal-Lawson
The core here is-
Per Axbom
There’s no evidence, no evidence for that assertion, either.
James Royal-Lawson
The core here is, he said, however, your design will alienate normal visioned users who make up the majority of your user base. I mean, he just goes, he, all this goes completely against the whole idea of inclusive design, and promotes the idea that you should, you should only ever focus on the majority, really. Because even though, he, under the surface is saying, “Oh, there is a bit of a compromise.” It’s a compromise that’s just, at best, what he’s promoting is a compromise that’s just a little bit of recognition to the minority. Well, we’re not gonna really worry about them.
Per Axbom
Exactly. And here’s, here’s the thing, then you can’t, you can’t even compare those two needs. Because if you have a preference for a certain aesthetic, that’s very different from having a legitimate access need. If it’s not aesthetically pleasing, but you can still use it, and more people can use it, that’s more legitimate than designing for aesthetically pleasing and less people can use it. It just doesn’t make any sense. Because, I also have, I reacted to, to early on also, this phrase: the goal of accessibility is to meet the needs of the minority because they’re often forgotten. How is that the goal of accessibility? The goal of accessibility is to understand when you see every one of us as human beings with a right to information, to understand how you accommodate those people, or the most number of people, it’s not because they’re often forgotten, it’s because we see them as human beings.
James Royal-Lawson
Yeah.
Per Axbom
It’s just, it’s, that’s, and that, I think that’s what makes it emotional for me because there are so many things in this that actually, almost are offensive to a lot of people. But if you assert that, you are called an extremist. You are called an accessibility extremist or accessibility snowflake-
James Royal-Lawson
Yes.
Per Axbom
– as the original tweet said.
James Royal-Lawson
Yeah.
Per Axbom
And, of course it is, it is intended to provoke. That’s how I feel about this article. And even so, and, and the problem with these types of articles, of course, then, there’s always something in it that is correct, yes? Strive for AA, AAA is not always what you strive for, but also you can have the high contrast can be behind another button. I don’t always approve of that. But it’s some, some, some statements in the web content accessibility guidance could be interpreted that way, that, that’s a good way to, to make a high contrast high contrast version behind another buttons. But, so there are some things that are right, but it’s presented in a way that the article itself actually misleads and alienates people.
James Royal-Lawson
And I’d go a little bit further to say that, I mean, I completely agree with you that this, this article, clearly is trying to provoke people. But I, I don’t think it’s just provoking. I think it’s dangerous, because of exactly what you’re saying that it’s painting out things. It is laying things out, to many people as a reasoned argument, and on the surface, some of these things come out believable and correct, you could argue; but, it does such a disservice to so many aspects of how we, we should be working. And I think- Interesting, interesting suggests that there is an aesthetic accessibility paradox. I mean, you, they aren’t two things that are in constant battle with each other. They are, they are not; they, they shouldn’t be. They don’t have to be. And if you put them up against each other, then you’re doing something wrong. And this article is wrong.
I mean, right. I mean, it does. Some of the comments go on about there being a trade-off; that we’re always in situations where we have to make design trade-offs. And that is actually true. What we, what we do is hard. The business environments or the organizations we work in are often very challenging. And, lots of people have very strong opinions. And, you know, we, you don’t, you don’t get your, you don’t get your point across at every, every corner, you don’t even maybe have all the knowledge that you need all the points; you know, we get stuff wrong, we need to do stuff again, and so on. So you will make trade-offs. But it’d be really good if you do make a trade-off. Then, you do it in situations where you’re happy with it and you’ve made a conscious, informed decision.
So you understand- You may have- the harm that you could be causing.
Per Axbom
Exactly. Understanding that is key, documenting it is key. Being able to motivate that decision to others is key- all of that. I actually, I had a quote I want to read from the comments from a colour who, colorblind person: “This article makes me feel like I am the problem for being colorblind,” and I really understand that, how, where he’s coming from when he says that. And I have another quote from another accessibility article that I’ll put in the show notes. “Accessibility drives aesthetics because the ability to use something is beautiful.”
James Royal-Lawson
Oh.
Per Axbom
And I just love that quote. Because that’s what it’s all about.
James Royal-Lawson
Oh, that’s lovely. Yeah.
Per Axbom
Yeah.
James Royal-Lawson
That’s actually probably a good point to wrap-up on the first article.
Per Axbom
It probably is.
[Music]
Per Axbom
So moving on: “Three Principles for Designing ML-Powered Products.”
James Royal-Lawson
ML?
Per Axbom
And if you don’t know what ML, exactly, yeah- machine learning powered products and kudos to the authors here for not just throwing out the phrase AI. I love that.
James Royal-Lawson
Yeah.
Per Axbom
But also, not so good putting that abbreviation there.
James Royal-Lawson
No, they explained it in the first sentence. So.
Per Axbom
Well, yeah, they do in the first sentence. You’re right. Yeah.
James Royal-Lawson
I’m not complaining about it. But it’s my turn, Budelman. And, how did you pronounce this Mark surname?
Per Axbom
I’m saying kiss-el-shteen or kiss-el-shtyne.
James Royal-Lawson
Kiss-el-stine. We could do a, we should do a separate podcast at some point with learn, Teach James How to Pronounce Names.
Per Axbom
Exactly.
James Royal-Lawson
But they’re both, they both worked for Spotify Design. And, the article goes through the design process behind some of the machine learning driven features of Spotify, such as the Discover Weekly playlist, and the start page, this, the home section of the Spotify app.
Per Axbom
And, their first point-
James Royal-Lawson
I do want- should we list all three straight off so, so people kind of like-
Per Axbom
Do it.
James Royal-Lawson
Okay, so the three, so these, these three principles for designing machine learning powered products, according to Spotify Design are: identify friction and automate it away, ask the right questions-
Per Axbom
Number two.
James Royal-Lawson
And, go manual before you go magical.
Per Axbom
Which, on their own, maybe, won’t say so much; actually have to read on to understand what each point is about. And, you told me you would react, I would react to the first point. And of course, I did, because it uses the word friction, which is one of my pet peeves; and, identify friction and automate it away because every designer understands that we always have to reduce friction as much, as much as possible. And, but of course, listeners to this show will know that I argue a lot for the addition of friction where it’s relevant to actually help people make better decisions.
James Royal-Lawson
Well also not just the, the addition of friction part. I mean, I think we argue, we’ve argued over the years, but just that whole thing of checking about the need for friction, when you sometimes you-
Per Axbom
Exactly.
James Royal-Lawson
Remove, rather, checking for the need of removing it, but you think about adding it, which is also a valid point. But you also kind of really have to think what happens when I remove friction.
Per Axbom
Yeah, but exactly. But the friction has a place; it has its own… There’s a reason we, friction helps us drive on roads, because otherwise we would slip off.
James Royal-Lawson
Yes.
Per Axbom
So, that that makes sense. In all, it’s all actually, all professions and industries.
Per Axbom
So that’s why I actually object to the example of reducing friction, because a better example of reducing friction is about instead of waiting two minutes to download a specific song, Spotify users could immediately play any song anytime, anywhere. And that is sort of the success story of Spotify, but they actually were able to accomplish that much earlier than anyone else could do it online. And of course, that is what Spotify is all about; but I don’t listen to music.
James Royal-Lawson
Yeah.
James Royal-Lawson
Kind of, but I, when I read this, I get kind of- my brain starts wandering off and I’m thinking, “God if that’s if that’s Spotify’s goal, to kind of, Spotify’ll play music, anytime, anywhere, removing all friction, then it would just start blood. It just blaring out music whenever I came near the Spotify app, I wouldn’t even need to touch anything. It would just be constantly screaming music at me.” That wouldn’t be a good experience. And I think that is more the point they’re trying to make. I think they’re just, they’re just glossed over that aspect of it, that they’re, they, they, they do really want to have a good experience. So they, they are-
Per Axbom
Exactly.
James Royal-Lawson
– promoting that you, you find friction in the form of things that are hindering, in a negative way, what you want to achieve with the app. And-
Per Axbom
Exactly.
James Royal-Lawson
Trying to remove it.
Per Axbom
They have friction because they actually have, they have buttons to control the music. And they actually do have pauses between each play track. So there is friction that they have, they just don’t, maybe not, acknowledge it as friction always. But I’m not being entirely fair either, because they actually do define, they define friction as anywhere in the user experience where human struggles in pursuit of their goals. So, you actually have to figure out the goals first before you decide to reduce friction.
Per Axbom
Yeah. And ultimately, you’d want to check to make sure those goals are actually healthy goals, or fair goals, because you don’t want to necessarily support all users in all their goals.
Per Axbom
But I also wonder sometimes when I was thinking about this when I was reading about Discover Weekly, and it’s not something I use, but I understand that people really appreciate it. But doesn’t that, if you always use Discover Weekly, have always given the, the power to someone else to decide for you what songs you’re, you’re going to discover, you never actually sit down and, and put that effort into discovering music yourself, you just sort of rely on that machine learning process to do it for you.
James Royal-Lawson
A little bit. Although of course it needs the, the songs you play from Discover Weekly are not used to generate Discover Weekly. So you’ve got to, you’ve got to listen to stuff outside of that playlist to allow that playlist to generat it effectively.
Per Axbom
And that’s, that’s an excellent point because that actually, that actually brings us into what they’re saying at the end here. It’s, it’s a win, and what I really like about it is that it’s a combination of augmentation automation; you have to decide what to automate. And you have to decide what you augment. And also, of course, what still has to stay manual. So that balance always has to be there. And I liked that they acknowledge that.
James Royal-Lawson
Yeah, exactly. This is one of the main points about this, that it’s identify friction automate it away, when it’s suitable to automate it, and don’t automate it away completely. So they really are pushing the thing about enhancement and augmentation and, and reducing friction. So they are actually, I think, tying into what you said about they’re not trying to remove friction completely. They are just, they are just really trying to make a better user experience, which is itself. Yeah, so as far as as far as the first principle of designing machine learning powered product, this is actually a pretty good principle. So we’re agreeing with the principle.
Per Axbom
Absolutely.
James Royal-Lawson
I think similar thing for number for the second principle: ask the right questions. Now they start off this section by by quoting, Picasso, “Computers are useless. They can only give you answers.” Which is a wonderfully philosophic, philosophical statement. You can’t give the right answer unless you’ve understood the question. And, that’s essential with machine learning too.
Per Axbom
I like that, yeah.
James Royal-Lawson
You, you, when you’re designing something based on machine learning, how can you get the computer to give the right answer unless you’ve really understood what’s going on? So, this, this principle shows how Spotify’s tried to embrace that when designing – use this to help shape algorithms in human-centred ways. And try to focus on the human side of this first before they dive into the machine side of things. They write here: “What does it mean to like an artist, or an album, or playlist, or podcast on Spotify? How does the user’s context shape the decision of what to listen to? What does someone need to know before making the choice of what to listen to?” So, so by asking these questions, or even hundreds of questions, it forces them to put the actual user scenarios and user needs at the centre of the algorithm.
Per Axbom
Yep. With one caveat, when I was reading this, because I was always I wasn’t playing devil’s advocate when I’m reading these articles. At first, when I was looking at it, why do they keep asking these questions to themselves? Why don’t they go and ask them to users, but when you read on, they start listing the questions and to find the answers to these questions. They actually are they have to go out and talk to people. And they used the assess the user feedback and identify bayla patterns in the data. So they’re both looking at qualitative and quantitative data to understand this, and it was only after they Started proving the hypothesis they had based on these questions that they started to apply machine learning. So I like how they’re describing that. We move very slowly, actually. That is sort of I’m getting with this, we move slowly and consistently and with intent before we applying the complex areas of the design,
James Royal-Lawson
yeah, you getting in now to the third principle that they list “go manual before you go magical”.
Per Axbom
Right.
James Royal-Lawson
This is, they say here as Spotify designers. You get certain stage during product development and typical delivery will be a wireframe or prototype of the intended experience. But when designing machine learning experiences, it might look like a set, it might be a set of rules that you you deliver. You get the question of like, how do you prototype a personal experience is what there’s what they bring up in the article, the challenge of that, when you’ve got to this point, which they say well, you’re maybe not going to go straight on to using the data first. So they use hypotheses and start to manually build up rules that might succeed in fulfilling that hypothesis. And once they’ve got a set of rules or heuristics, as they write in the article, they will actually maybe get the database to build something that pulls out the data to generate that. But I’ll quote another bit from the article here, “if we can’t achieve a good quality experience with a manual approach, applying machine learning to the problem, probably won’t be worth the investment.”
Per Axbom
Exactly.
James Royal-Lawson
So So if they build up they build up their heuristics on their questions and hypotheses, and then try and answer them themselves. And see whether this kind of constraints or things that would work and only one they’re satisfied with that kind of self check. That they will then probably Go forward to putting effort into setting the algorithm loose.
Per Axbom
Exactly. I really love how this – I’m realising now that this article is the is really explaining when when people say, just because you can build it doesn’t mean you should build it. And when if you wonder what that means you can read this article because this is what it means it means that you are really intentful about finding out if you should build it. I’m liking it more and more as we’re discussing it actually because I found a thought I found some faults in it, but I really like what where they’re coming from they’re talking about speed in the end of the day that they moving as fast as possible. But I I really like how slowly they’re moving as well.
James Royal-Lawson
Yes!
Per Axbom
because that’s what I’m taking away from it.
James Royal-Lawson
This is this is this to me. I mean, we we started by focusing on the bit the friction side and the danger of that, but yes, absolutely, ultimately, and they themselves say the techniques that they’ve describing in the article. These are all techniques derived from long established human centred design principles. And, and it’s true and you can tell that we’re in the whole article. And and I love the fact that they’ve really shown the, the benefits and the advantage of investing quality time, quality research to truly understand the problem, create a working solution. This this is the kind of approach I think we lack in so many aspects of it, especially when we’re forced with, with with the various methods of working that we use with you know, agile and, and sprints and so on where we were, and skateboards for what’s it called – minimum viable products, all these things that push us to do stuff quickly and get stuff out there. Whereas here we’re seeing the, the the benefit and the quality that you can get, and ultimately the better user experience you can deliver. By investing genuinely in the design.
Per Axbom
I’m wondering now if this actually is an effect of machine learning being quite complex and expensive. So that you realise that if you invest too much and the wrong solution, it will actually be very expensive in the long term, or also in the short term as well. Whereas, during design, during the 2000s, and 2010s, we’ve been we’ve been we have had so many designs, that are so cheap to produce. So you’re actually you can go two weeks and you all you have an idea and another two weeks you have it published out there. And that’s sort of what’s creating a created a lot of the problems we have in design today where you release solutions that you haven’t really checked What harm they could probably could possibly create. But in this case, you actually have something that you have to create that is quite expensive and difficult to maintain. And so you have to be really, really sure that you’re doing the right thing, which actually forces you to be this careful about what you’re doing.
James Royal-Lawson
I think that’s a good point about design being cheap. I mean, we many times you’re doing multiple variations of designs and A B testing them and so on. Because it’s so cheap, you can actually throw out two them.
Per Axbom
Yeah.
James Royal-Lawson
Or three of them or whatever. No. So I mean, they, they do work into it iteratively, they say in it, that they do many rounds of some of the manual part of this design. So it’s not like they do one designed straight away and they throw it out. They really do take their time to, to iterate and to test, to see what’s happening. Is it achieving what they want, before taking that final one or final that that next big leap into production, into release, to a large user base? Yeah. So I think we can definitely learn a fair bit about healthy design from, not just for machine learning, but design in general, from this article about
Per Axbom
Yeah
James Royal-Lawson
machine learning from Spotify.
Per Axbom
I’m browsing ahead in my notes here and i’m i’m looking at what episode we were supposed to recommend for the show. And it’s it’s it’s fantastic because you actually found an episode from way way back: Episode 66
James Royal-Lawson
Which was 6 years ago!
Per Axbom
Yeah. Also an article by Anthony Tsang, was that also on UX movement, I guess?
James Royal-Lawson
Yes.
Per Axbom
Yeah. Okay.
James Royal-Lawson
It’s a classic episode.
Per Axbom
Is it the white rabbits?
James Royal-Lawson
Yes.
Per Axbom
Oh my god.
James Royal-Lawson
Of course, it’s the epic delete undo argument and It’s the peanut butter and jam episode. Where Jared spool and sorry, Peter
Per Axbom
Merholz.
James Royal-Lawson
Yes. argue about whether UX basically destroyed IA – information architecture.
Per Axbom
Wow.
James Royal-Lawson
Peanut Butter jam comes up as the as the the Jared-analogy. It’s an excellent thread. It’s a classic episode.
Per Axbom
Haha. And the epic delete undo argument is really the first and only time that you and I really argued.
James Royal-Lawson
Pretty much. Yes What UX movement does to you hope you enjoyed today’s linkshow and if you’d like to contribute to funding us podcast, please visit us podcast.com/support.
Per Axbom
Remember to keep moving
James Royal-Lawson
See you on the other side.
[Music]
Per Axbom
Why was King Arthur’s army too tired to fight?
James Royal-Lawson
I don’t know why was King Arthur’s army too tired to fight?
Per Axbom
It had too many sleepless nights
This is a transcript of a conversation between James Royal-Lawson and Per Axbom recorded in January 2020 and published as Episode 228 of UX Podcast.