Let’s dig in.

What are emergent abilities?

In the 2022 paper: “Emergent Abilities of Large Language Models” researchers at DeepMind are the first to detect a weird phenomenon that they describe as a “sudden appearance of novel behaviour” in their LLMs. Since the release of that paper it has now been detected in most AI models, and always seem to happen as they reach a certain volume of training data. At this point the models all seem to “unlock” knowledge that is unrelated to what they were trained for, but exactly when this happens and what this capability is, has so far been unpredictable.

Examples

Models that are not trained on math has about a 0% chance of answering math related questions. The same can be said for linguistics, physics and so on, as illustrated below:

This is entirely expected. But as the model size continues to increase, there comes a certain point where the AI is suddenly able to answer questions related to these domains that are specifically NOT part of their training data:

What’s even more interesting is that it doesn’t just gradually start to understand a few basic concepts within those new domains. In most cases it has been a very drastic increase, that clearly breaks any previous trend within that domain.

This means that the notion we had of AI’s being mere “mimicking machines” is factually wrong (as they reach a certain size), at which point there’s something more than just imitation going on.

Guess the movie

By now there have been countless observations of this effect, and the author of the DeepMind paper that started it all has this blog where he keeps track of them: https://www.jasonwei.net/blog/emergence
At the time of writing there are 137 examples that have been empirically discovered, some on purpose, but most on accident.

A fun example I found was a model that apparently unlocked the ability to correctly guess movies based on emojis. Also known as the “emoji quiz”, (a popular game online), the goal is to guess which movie the emojis represent, like these:

As this was not part of the AI’s training set, it naturally got about 0% of these right, and the riddle-like format of the quiz actually makes it quite tricky for LLMs to solve (some of the examples are even pretty far fetched). But as with the other discoveries of emergent abilities, the LLM got to a point where it cracked the code and from that moment, it didn’t just get a few right, it was able to answer these at human level accuracy, despite not having seen this quiz or (even similar games) in its training data.

What is going on?

Well, no one has definitely answered that yet. A common theory right now is that underlying patterns in data actually do overlap, we just don’t notice it. For example an AI that was trained for translating between a few dozen languages suddenly understood languages it hadn’t been trained on too. As we know, languages are based on rules, and anyone who has tried to learn multiple languages can tell you that there are patterns you start to notice, and suddenly you too become pretty good at guessing what words in a foreign language mean, because you see similarities to the languages you know.

Similarly it is speculated that there are a lot of patterns and structures in everything we read, see and watch that we are just oblivious to. In some sense you could compare it to human intuition. You can’t quite explain how you arrive at your answer, but based on all your different experiences and everything else you know, you just have a feeling.

What does the future hold?

It would be exciting if LLMs in the future are able to articulate how they arrive at their answer in detail. It isn’t difficult to imagine that we might discover how vastly difference disciplines actually consists of the same underlying concepts, and that uncovering these will lead to new ideas and innovations. However it all starts with us finding out if this theory is even true, and if so, how we can detect these fundamental connections. I would bet that when we find the answer to that, it won’t just impact how we develop future LLMs, but also lead to new breakthroughs in many scientific disciplines.

Tip of the iceberg?

As we continue to increase the size of our models there’s a chance we will discover another point in larger altitudes of data that also exponentially increases the model’s capacity. GPT4 is trained on about 1 petabyte of data, and the human brain averages about 2,5 petabyte of data (although these are clearly difficult to directly compare). This is pure speculation, but the research shows that emergent abilities are most often unlocked as the model size gets to 0,5 petabyte of data. This is not a magic number, as some models triggered emergent abilities at 0,4 and 0,6 petabyte of data. However there do seem to be a correlation around this area. Now imagine that we increase our models to 2, 3 or 4 petabyte of data. We could in theory find that even more powerful emergent abilities happen at some specific points on this curve. And that these emergent abilities completely breaks the predictability of our model and instead makes 10x “smarter” than intended (including knowledge on topics we deliberately leave out of the training data).

Shouldn’t this get more attention?

In the constant flood of AI news I can see why this one didn’t reach the front pages as it’s still a bit technical, and people don’t like when something is inexplainable. In fact, it might be in the interest of AI companies to not talk about this phenomenon too much, as it would be easy for the media to spin this as AI’s already being out of control.

To me this is the most “underrated” AI discovery we’ve made so far, and even though I make it sound like a big deal, there’s of course also the chance that it’s a singular occurrence at these specific sizes of models, and that we won’t find it again at higher levels. In short, we just don’t know much about this right now, and that makes it very difficult to predict how big of an impact this discovery will have in the future. There’s currently much more research ongoing, but I’d argue not nearly as much as there should be. There’s a good amount of papers already that try to find causes and explanations for this phenomenon, suggesting that something else must be going on, and while that could definitely be true, none of them have found it yet. At the pace of AI right now, I don’t doubt for a second that this article will be irrelevant in a few months. At least I hope so, and I hope you will now share the excitement with me too as it does, and we learn what’s really going on here and how deep the rabbit hole goes :)

Sources:

Original paper (from DeepMind)
https://arxiv.org/abs/2206.07682

Jason Wei blog (main author of the DeepMind paper)
https://www.jasonwei.net/blog/emergence

Criticism of emergent abilities
https://arxiv.org/abs/2304.15004

Thorough summary (and source of the emoji quiz example I mentioned)
https://www.assemblyai.com/blog/emergent-abilities-of-large-language-models/

Newest research paper on the study of emergent abilities:
https://arxiv.org/abs/2206.04615

Why ask this: It’s a clear red flag if they don’t know or simply do not measure a product’s success. But remember that for smaller companies this might be exactly why they want to hire you. If you want the opportunity to build how they could measure this in the future, this red flag might be a blessing in disguise. But if you’re looking for a more mature design organisation, this is probably the first sign that you should continue looking elsewhere.

How to interpret their response: If the company measures, NPS (Net Promoter Score), customer satisfaction and retention rates, then you know these are user-centric metrics.​

“Do you have budget for qualitative research?”

Why ask this: Most companies know that customer feedback is valuable, but probed further, many companies tend to neglect it regardless. Many will mention that ‘one time where they actually did it’ but try to dig deeper by asking if they set aside budget for research on an annual basis and research is consistent throughout their products.

How to interpret their response: A key indicator of a business-focused company may be that they use market research over their own individual research.​

“On your last feature / release did you do any evaluations before release?”

Why ask this: Just like interviewers ask about your most recent project, asking them about their most recent release is important too.

How to interpret their response: It might not be the one they are most proud of, but that’s exactly why you’re asking this specific question. You want a realistic picture of what you can expect, not just stories from their greatest hits.​

“When deciding what to build next, how are user insights weighted?”

Why ask this: If you’re applying for a more senior position, then this matters more. You don’t want to just test what others have decided to build, you want to have a seat at the table and steer the direction of the roadmap based on actual research and insights.

How to interpret their response: User-driven companies will be responding to direct customer feedback and see this as invaluable to improve whereas business-focused companies may be a little more selective with the input from customers.​

“Have you ever decided to postpone or entirely cancel a feature based on user tests?”

Why ask this: This is normally where companies struggle the most. Even if prototype testing and user evaluations show that this might turn into a disaster, there’s simply no going back now that they’ve spent months working on this feature. …Or is there?

How to interpret their response: This is where a company can truly shine, if they can highlight a release where senior management decided to change the release date or the scope of the release based on input from designers then see this is a green flag.​

“How does design impact translate to compensation in your company?”

Why ask this: You want to know if the company offers a bonus or other incentives based on performance because this will directly impact your role as a designer, more specifically the balance between the user and business needs, as well as stakeholder management on a project.

In situations where a big part of your salary is based on performance, the metrics that the employer use to evaluate impact must account for customer satisfaction too.

This is almost non-negotiable, simply because if it is not, you won’t be able to properly do your job. If the only incentives for your role are to onboard as many users as possible and make them pay as much as possible, there is nothing stopping you from going down the dark pattern rabbit hole.

This can end up hurting the company eventually, however, it is important to strike a balance. As a designer you are the voice of the users, but you also need the company to stay profitable and grow.This is in my opinion where the most exciting part of our responsibility as designers lie.

How to interpret their response: It is important to see how the company values your contributions as a designer, and you are ideally looking for companies that have some form of customer satisfaction as part of their compensation scheme.

Note: this article is a draft that I continuously work on.

Why should UX designers care about Blockchain?
Blockchain became the buzz of 2017 not because of the amazing potential of the technology, but as an investment object. Now in 2018, the huge gains that used to surround cryptocurrencies is gone, left stands an incredible technology with almost no adoption rate — yet. As was the case with the internet in the 1990’s, the technology to change the world as we know it is there, the nerds who will build it are present, so what’s halting us? Well, nothing will happen as long as regular people won’t use it. In the case of the internet, it took more than a decade to reach mainstream adoption. Three things helped that happen.

1. Speed increased, due to better hardware and internet connections.
2. Use-cases & content increased, due to a longer timespan of trial and error, with creators figuring out how to make money by providing value online.
3. Lastly, Usability increased. This wasn’t what led the adoption. In fact, most websites in the 90’s had exceptionally poor usability. But to reach more users, and to beat competition, usability became the most important way to stand out in the 00’s and the following decades. As usability became a prioritization, mainstream adoption increased.

What we see now is that Blockchain developers all over the world are already hard at work on the first two criteria, but not a lot of attention is given to increasing the usability of the blockchain technology. In the 90’s online usability didn’t just happen. It was led by pioneers like Donald Norman and Jakob Nielsen. Now we are in a somewhat similar place, in the way that we have a technology with huge possibilities, and with only very few designers, caring about how to present it in a consistent and simple way, to achieve intuitive blockchain experiences.

Having worked on a few blockchain projects as UX designer for the Nordic Blockchain Association, these are some of my tips for fellow ux and ui designers getting into the fascinating world of blockchain technology.

Does blockchain need new design principles?
To begin with we need to break this down from a design perspective.
The internet was being designed in tandem with designing for screens, for interfaces and GUI’s in general. This is not the case with blockchain. Today we have a lot of frameworks, best practice examples, and ideas of what works and doesn’t work in cyberspace. Most of this can be directly applied in the same way as we would with an app, a website or any type of software. The same design principles that goes all the way back to Norman and Gibson can be used on blockchain dApps without any adaptation, since it was the same underlying foundation from a design perspective.

Why aren’t more designers looking into this?
Well, one of the reasons is that most designers see it as too complicated to get into and they wouldn’t know where to start. Unlike AR and VR that has specific apps (like Lens Studio and Unity), Blockchain is first and foremost a technology that lies in the developers realm. The magic of blockchain happens under the hood, not in the interface. Many interfaces that we are used to, Financial services, games, advertising, and analytics, just to name a few, are also being built on blockchains. If you have experience in any of these areas, you shouldn’t hold yourself back from applying for a job in a blockchain startup looking to put that service on the blockchain. In fact, it’s probably better for you as a designer to know as little as possible about blockchain. This will make it easier for you to ask the right questions, and empathise with potential users who are also new to the world of blockchain.

Current problems no one has solved yet
We need a layer like the Ethernet Network Service. People don’t remember 16 digit bank accounts and they don’t remember the IP addresses for all the websites they visit. You cannot remember Google’s IP address because… why should you? You can just type in the name! What DNS did for IP adresses, we need for blockchain addresses. It would be easy to counter this argument by saying I can’t remember the 16 digits on my credit card either. But even though this is what we are used to, wouldn’t it be smarter to have a regular name for this code that you can remember? Well yes and no. Easier for you, but less secure right. But this is not the case for crypto. We can use the advantage of public and private keys. Whereas it would only require you to remember a date and three digits if the main card number was easily memorable, this is not the case in blockchain. The private key should stay as impossible to remember as it is, but no reason that the public key is just as impossible to remember. Donating will be as easy as remembering something like Save-the-animals, 1–800-WWF and so on. but what if I make an address that is Safe-the-animals? Like people making gogle.com and scam sites like that? That’s the biggest issue that I don’t have a solution to yet. If these addresses are impossible to remember, people will HAVE to copy/paste it, ensuring that you don’t mistakenly type in a wrong character. At the same time someone could fake being, say, Elon Musk on Twitter and ask for donations to a bitcoin address. How can you say if it is his address or not? It’s just a bunch of random characters. Banks do this by giving some, but not all information about the address. When I put in an address in my bank’s website, it tells me which city and bank it belongs to. This is still not ideal, since I could create a fake account at the same bank, which means I wouldn’t notice any difference. But in Blockchain we can change this paradigm too. What if each address had a unique word or sentence? When I put in Elon Musk’s address I know it should say “Mars 2020” or whatever message that Elon has permanently linked to this address? In Denmark this is pretty close to mobile pay. You put in a phone number that you don’t know, and ensure that you typed it right by looking at the name that shows up linked to that number. Each user on the blockchain could choose to use either their real name or a statement, based on whether they value their anonymity or not. My friend sends me his address, I confirm with him that his word is “Catfish”. I scan a QR code at the table of a restaurant and confirm with them that their word is “Epi Tempura”, to ensure no one replaced the QR code with their own. I show that I click transfer to the restaurant owner when I pay, and I leave. We don’t have to wait for the transfer to go through, using the non-reversible nature of the blockchain for a smoother restaurant experience.

The company Webmoney has their own take on this. They have implemented an optional protection code. Basically a codeword that you put in together with your transaction. The receiver will have to type in the same codeword to receive the money, if they fail to do that within 3 days the order will revert back. This is a less smooth version of my above concept, and it has it’s own flaws. In stores I could cheat by replacing the O with an 0 in my protection code (or something similar). The owner will not be able to get the protection code right and I will get my bill refunded.

Even though it is not a popular opinion in the space, I think that blockchain will not overtake the banks but coexist together. There will be those who want to have full control of their own money, and those who want a bank to take care of it. For elderly people the latter would be recommendable. It will be like asking what bank you have, some will have their own wallets, others will have traditional banks. UX doesn’t have to make blockchain the default for everyone. Another unpopular opinion. But it is setting the field up for failure if this is our mission. It should be accessible to anyone who wants to, but shouldn’t aim to move 100% of the population on the blockchain. The elderly people I know make tons of mistakes on their smart phones that I have to solve. This is ok, it’s easy to solve, no harm done. This is not the case in blockchain. I would not be able to get their money back if they accidentally sent it to the wrong address. A very simple feature that is currently missing is the ability to save blockchain addresses as contacts and name them properly. This could make it easier and less cognition intensive to transfer money between your own accounts and your friends, companies or other addresses that you might need often.

Blockchain Design systems
When the main offering of blockchain is trust, it’s interesting how mainstream audiences find the technology interesting but doesn’t trust it because they don’t understand it.

Connie Yang used to be product designer at Facebook but is now working in blockchain and agrees that fundamentally UX philosophies still apply just as much in blockchain companies.

Monaco onboarding flow?

UX Principles that are currently undervalued in the blockchain community:

Minimizing cognitive load. It takes a lot of cognitive powers, and in part because it’s all so new to us. I think using the internet in the 90’s required way more cognitive power too compared to today, because we just weren’t as used to it.

Optimize user flow. I have tried out so many dApps that I’ve lost track. But there isn’t one that I keep going back to, and none of them have asked me why. Apparently no one cares what I as a user has to say about my experience. 1. analyze the data. Where do users jump off and never come back? 2. If it isn’t obvious why (and actually even if you think it is obvious why) ask your users about what lead them to the conclusion that this wasn’t as interesting as they thought. For me personally it has mostly been a) I didn’t understand what I was supposed to do next, or b) I didn’t find the content or functionality that I expected. If these companies asked me what I expected, or what I didn’t understand, and managed to fix this, they would already be way ahead of competition, and only then will user growth start to happen organically. Allow all users to give system feedback, and don’t dismiss the feedback as “that’s too difficult to fix”. If it can’t be fixed, maybe there are ways to explain it or make it less of a problem to the users.

Remember to research your non-blockchain competitors. This is not a blockchain-only principle, every company has been trying to optimize this for ages, and the succesful ones have often found clever solutions. Find best practice solutions to your problems, chances are they are better than what you currently have. Steal them, adapt them.

On boarding

Like with any platform, onboarding is very important to make as smooth as possible. Make interactive tutorials, make it easily learnable and help the user get going with getting value out of your product before you leave them to it.

Where differs blockchain from other online experiences?

The cost of human error.
Despite the system is way more autonomous than traditional banking, the risk and cost of human error is much higher. There is no undo button. This only relates specifically to transactions though, but is still an important aspect that no one has addressed yet. The right amount of warning messages (not to many for users to start dismissing them) could help make fewer mistakes, and please explain these warnings about what is happening in an accurate and understandable way. We could also analyze what it is users do most, and where they most often make mistakes. Empower the users to be in control over these irreversible actions.

The response time limits
https://www.nngroup.com/articles/response-times-3-important-limits/

“you should send transactional email and confirmation messages within 1 minute to keep users from wondering whether their action — such as a purchase or address change — has been received correctly.”

“The difference here is that users assume that customer service requires human intervention, so they don’t think the computer is broken if they don’t hear back within 1 minute.”

Selvom der arbejdes på at dette bliver hurtigere rent teknisk (Det er et af de områder som der arbejdes allermest på i øjeblikket, alt fra Lightning network til IOTA til Proof of Stake teknologier søger at gøre transaktionshastighederne hurtigere) så kan UX formindske problemet ved at bekræfte brugeren i at det er helt normalt og forklare hvad der sker.

https://www.nngroup.com/articles/powers-of-10-time-scales-in-ux/

Brug billedet. Forklar hvorfor det tager tid. “Netværket er nu ved at godkende din handel, dette kan tage op til en dag”. Show a progress bar if possible. It could be an entertaining animation of how people or anthropomorphic computers are right now arguing over whether this trade is legit.

Volatility

If I were to purchase tokens for whatever decentralized platform that I’m interested in. It could be the decentralized clone of Udemy called x. This platform requires you to acquire it’s tokens to purchase the tutorials. Now the first issue here is that I need to create an account on an exchange that trades this token first. Of course this breaks the user flow. Playstation store has the same in principle, but it accepts all currencies ON the platform to purchase tokens for that you can then buy content with. Number one would be working towards letting the user stay on the platform without having to sign up at other pages to access your content. Number two, the value of this currency should not be volatile. It seems many platforms in crypto want users to purchase their tokens, not to spend on the actual products they offer, but because they will increase in value over time. In theory it should incentivize adoption rate, as sort of an early bird discount. The sooner I get in, the more value I get for my dollars. But in reality it makes the value so volatile that the opposite can happen (and has happened). This is an interesting dynamic, as I’m not interested in buying 100 tokens for $100 when I can risk that the next week I would have gotten 1000 tokens for $100. But at the same time I might buy it the week after if it seems like I will get a good deal out of it. The volatility of tokens works just like offers and discounts. A day where all tokens in the market drop in value is the equivalent of black friday (or cyber monday if you will). This is a way to explain the dynamic in a positive light and relating it to something users already know. Now it’s not as frightening that the token is volatile, and even though we are reaching the domain of copywriters (or UX copy), this is also an important aspect of gaining the users trust by explaining what’s happening.

Questions a UX’er should ask when they begin work on a blockchain project:

Why would users want to join and stay on this network? Are there any incentives for users to stay on the platform/get in early/contribute/invite their friends/etc. Take notice of the incentives (or lack of incentives) and think of whether this is the most optimal way from a user perspective. Ask yourself what model you would prefer, and ask the potential users what model they would prefer.

What are the user stories needed? As a “member” I want to “action” because “need”.

This is a way for developers (and everyone on the team) to agree what we are building towards, in a language that is non-technical.

What non-blockchain company is our closest competitor? At this time in history, blockchain based companies will not be competing with other blockchain based companies for the users. The users are enticed to stay at the very well-functioning platforms they already know and use, that can (in most cases) do more or less the same as the blockchain based alternatives. It’s your job to find out what arguments or values that this technology actually provides that will convince users (who are not pr. default impressed by knowing that the app runs on blockchain) to actually switch from what they know. It’s said that for users to switch it has to be 3/4x better (source? Can’t remember), if the platform/app isn’t there yet, make sure to lay out a roadmap for how they should get there. In order to actually understand what can be achieved and why the founders chose to make this app/platform you need to get your hands dirty and understand the fundamentals of how your product works. If you don’t get it, how will your users? The decentralized social media network x doesn’t compete with other decentralized Social networks. They compete with facebook, twitter and snapchat. Yes it’s great that it gives you gain full control over your own data, that might be sufficient to convince some, but I can guarantee that it will not be the argument that wins over the masses. Find out what painpoints you can fix that are so severe that users will invest time and effort in converting to your new platform instead.

White paper

The way most blockchain based companies start is by making a white paper where potential investors can read about the tech, the mission the vision, and anything else that is the secret sauce of the company to win mainstream acceptance. This is what could be labelled as the wireframe stage of the company. We outline what this is going to look like and feel like in the future. But unlike wireframes, the only ones that are looking at a white paper are the actual users. They are far too complex to understand, and generally not of interest for anyone outside of the developer and investor world. But this seems wrong from a design thinking perspective. This is the exactly the stage where we should do user testing, in order to not build a product that no user want. UX designers getting in at this stage should break it apart and adapt it a format that anyone could understand and tell you whether they would use it or not. Just as with wireframes it can be difficult for users to visualize the end product, so use whatever exercises and tricks that you normally would, to immerse the user into what this product is, in order to get honest and valuable feedback. Based on my own observations of 100+ blockchain platforms, most of these would have pivoted of they had had the chance to get this early feedback, into something that users actually need. It can even be profitable if it turns out the idea is so complex that no one gets it. Using this feedback to rewrite the white paper in a way that is easier to understand will make sure that potential investors get the idea, and might see why they should invest in this particular startup. Just mentioning that user tests at this stage showed a demand among users will increase the likelihood of someone believing that the project will succeed.

Keep it simple
Make sure to not overcomplicate anything. Even though the company might be very proud of the underlying Byzantine fault proof delegated Proof of Stake system that they have built, no user will care since they have no idea what that means. In fact it might scare them away. Being jargon-free is one of the core elements of building trust in the blockchain space according to IBM https://medium.com/design-ibm/blockchain-design-principles-599c5c067b6e

and therefore this should stay behind the scenes, in the same way that an email client wouldn’t explain how the underlying TCP/IP protocol works or a website explaining why they built it with react on https. Instead put focus on how the experience of using the product differs from what they are used to.

Replacing “forgot password”

One of the first things you will notice in blockchain is that you can’t recover forgotten passwords (including “private keys”). Telling users how important it is to store this password safely, and how they should do so seems to change a lot from platform to platform currently. We need to come up with a way to research what approach that works best, and unify this experience across platforms. Why not use Authenticators, like Authy or a blockchain alternative to save your private keys in? That way you will only have to remember/store the private key for that one app.

In case of Death
What happens in case of death? Right now the banks will hold the money till the rightful heirs are found. In the case of decentralized wallets it could lead to a finder’s keeper’s treasure hunt to find out where the private key is stored. Same goes for platforms such as decentralized social medias. If a relative dies you can let Facebook know and they will either deactivate or turn that profile into a mourning page, based on the request. When no one is in control of decentralized Social Media sites, who will be able to delete the profile? Will we have to make a consensus mechanism in the technology that can evaluate the claim of relatives that an individual is dead? We can’t just print more money, so if we are thinking long term, it could be a problem that more and more accounts are frozen because their private keys are never found. Will we eventually run out?

Summary
These are some of my initial thoughts about what is needed for blockchain to reach a mainstream audience. I hope you could find some use and inspiration from my observations about its current state, and I hope you will help me work out solutions to some of these challenges in 2019.

Det bør ændres, da der er mange andre fordele ved at lave løsninger med høj tilgængelighed, end at imødekomme den digitalt udfordrede del af befolkningen. Her er 6 mindre kendte grunde til at digital tilgængelighed skaber værdi for din forretning:

1. Øget markedsandel

Op mod 20% af den danske befolkning er digitalt udsatte, og har svært ved at bruge ikke-tilgængelige websites. Ved at lave en tilgængelig løsning, øges mængden af mennesker, der kan bruge din løsning, og dermed din markedsandel. Der er en tendens til at mange virksomheder undervurderer hvor stor en mængde af deres besøgende som har et handicap. Husk på at de fleste med nedsat syn eller førlighed i særlig grad er afhængige af at foretage deres handler og informationssøgning på internettet, da det kan være besværligt at forlade deres bolig på egen hånd.

2. Søgemaskineoptimering og øget trafik

Google, samt andre søgemaskiner, er rigtig glade for tilgængelige websites.
Et tilgængeligt website kræver, at du har styr på dine alt-tekster og title tags, at du har et sitemap, at du bruger dine headings tags korrekt, at du har beskrivende link- og billedtekster samt at din kode er semantisk. Enhver SEO-ekspert vil give dig samme anbefaling; Google belønner tilgængelige websites, og et tilgængeligt website betyder øget trafik.

I år lancerede Google en opdatering af deres design retningslinjer, og et af de afsnit som flest nye regler var accessibility. Google’s søgemaskine algoritme “straffer” i stigende grad websites som ikke lever op til deres retningslinjer, af hensyn til brugeroplevelsen.

3. Øget brugervenlighed hjælper alle

Et tilgængeligt website, betyder flere navigationsmuligheder, skarpere farvekontraster, tydeligere afkodning af ikoner, knapper, links, input-felter og billeder. UX-optimering af elementer som alle besøgende drager nytte af — ikke blot brugere med funktionsnedsættelse. De fleste brugere oplever nemlig temporær funktionsnedsættelse – altså at de i en given kontekst ikke kan tilgå sitet under optimale betingelser. Her er et par eksempler:

Kontrast
At sikre skarpere kontrast på teksten hjælper ikke kun brugere med synsnedsættelse, det kan også være en fordel for en bruger som besøger websitet via sin mobiltelefon i solskin nede på stranden, eller en bruger som benytter en gammel dårligt oplyst computerskærm.

Ikoner
Tydelig afkodning af ikoner og flere navigationsmuligheder hjælper ikke kun brugere med kognitive handicap. Ikoner skaber overskuelighed, og gør det lettere at danne sig et overblik. En bruger som kommer hjem efter en lang stresset arbejdsdag, eller et ældre menneske som generelt er utryg ved at navigere internettet, kan drage stor fordel af et intuitivt og let-overkommeligt website.

Input
Vi har alle prøvet at besøge et website på mobilen med for små knapper, links og inputfelter. At forbedre disse hjælper ikke kun brugere med et fysisk handicap, men kan også være til nytte for alle andre. For eksempel kan brugere med en lidt ødelagt mobilskærm, brugere som kun har én hånd tilgængelig, have særlige udfordringer ved disse. Det gælder ikke kun mobiltelefoner, der opstår også let situationer ved computere som komplicerer input-processen, hvad enten dette er en touchpad man ikke plejer at bruge, en for hurtig mus, eller en computer man låner med en uvant opsætning.

Et handicap behøver ikke være en permanent tilstand. Handicap afhænger i høj grad af vores mentale tilstand, fysiske kontekst og den hardware vi interagerer med. Vores situations-bestemte fysiske og mentale færdigheder, gør at vi alle agerer handicappede fra tid til anden, i vores daglige digitale interaktioner.

4. Øget konvertering
Regnestykket er ret simpelt.
Mere trafik + Lavere bounce-rate = Flere besøgende og øget konvertering.

En webshop med eksempelvis 10.000 køb hver måned, går glip af op imod 2.000 køb alene fordi brugerne ikke kan foretage deres køb grundet manglede tilgængeligheds-optimering. Hvis denne webshop forbedrer deres digitale tilgængelighed kan det potentielt øge den totale profit med op til 20%.

Bounce-rate
En bounce-rate er antallet af de brugere som øjeblikkeligt forlader websitet igen for at finde et alternativ. Ifølge et amerikansk studie er det ved dette stadie at det største frafald sker og op imod 71% af besøgende med et handicap forlader siden i dette stadie hvis de ser tegn på at siden ikke er tilgængelighed og dermed vil forårsage dem interaktionsproblemer.

Lettere navigation
Hvis navigationen af sidens produkter eller information er for besværligt og tidskrævende at gøre via en skærmoplæser, et keyboard (uden mus), eller måske for anstrengende for øjnene (grundet kontrast problemer eller for lille skrift), vil besøgende forlade siden igen uden at færdiggøre hvad de kom efter.

Købsflow
Sidst er der i selve ordreprocessen mulighed for at en vis mængde besøgende springer fra. Dette er også kendt som cart abandonment, og er kun et problem for webshops. Det kan være noget af det sværeste at designe, da der nemt opstår anledning til frustrationer i denne proces. Det skal være let og hurtigt at afkode og udfylde felterne. Ligeledes er rækkefølgen, placering, antallet, og validering af felterne vigtige parametre for en god salgs-brugerrejse.

Den største gevinst opnås når øget tilgængelighed sikres gennem hele brugerrejsen, fra brugeren lander, navigerer, og interagerer til de får løst deres behov og afslutter rejsen.

5. Performance og fremtidssikret kode
Semantisk korrekt kode er en måde at kategorisere sidens indhold, så det bliver lettere for både mennesker og maskiner at forstå det. Eksempelvis kan det hjælpe browseren med at se forskel på hvad der er overskrift og brødtekst. Dette hjælper blandt andet søgemaskiner med at præsentere det mest relevante indhold øverst i resultaterne. Inden for semantisk kodning er brugen af microdata en måde at gå et lag dybere i beskrivelserne.
Microdata kan indeholde tags som “person”, “place”, “event” og “review”, der hjælper webcrawlere og andre maskiner med at forstå konteksten af det indhold som vises. Dette øger i SEO-sammenhæng indholdets tilgængelighed på tværs af devices.

Microdata & WAI-ARIA
Ligesom microdata er WAI-ARIA en anden tilføjelse til semantisk kode, som sikrer at komplekse interaktioner bliver beskrevet i et sprog der forstås af computerne. Med WAI-ARIA kan computerne eksempelvis se forskel på primært og sekundært indhold, kalendere, dropdown menuer og andre interaktive elementer. Det øger graden af brugervenlighed, ikke kun for skærmlæsere, men også på tværs af browsere. Når computerne ved hvad det er for en type indhold som skal vises, er det mere sandsynligt at de præsenteres i det korrekte format. Dette sikrer også at ens digitale løsning er fremtidssikret til nye teknologier, browsere og browser-opdateringer, der i stigende grad kun understøtter sider der kan læses og forstås af maskinintelligens.

6. Øget kundetilfredshed & brand awareness
Et website som lykkes med at tilgodese de 20% af befolkningen der ofte møder udfordringer i deres færden på internettet, kan øge deres kundetilfredshed. Dermed har du et godt grundlag for at opbygge en base af loyale kunder, der har lyst til at vende tilbage til dit website. Da det endnu ikke er blevet en selvfølgelighed, giver det unikke muligheder for at generere positiv brand awareness at være frontløber på dette. For større virksomheder, kan indsatsen for øget tilgængelighed med fordel også inddrages i virksomhedens CSR-strategi.

Lær mere om tilgængelighed i WCAG’s guidelines eller kontakt mig for en snak om potentialet i netop din virksomhed.

Quick Summary

Anticipatory Design
Make sure your users opt-in, never force it on them.

Metadata
You really do need to have it to personalize, but if you don’t, consider using Nearest Neighbor technique.

Parameters
How do you get a user to trust your recommendations?

Control
Allow your user to tweak their experience.

Transparency
Show the user what inputs your recommendation engine use.

Trust
Is gained over time, and by continuously and consistently provide value.

Anticipatory Design needs to Ask for Permission

Buzzwords like anticipatory design have been circulating a lot, but research shows that users are still uncomfortable with having the service make choices on their behalf. The most important lesson is that users are ok with anticipatory design as long as they opt-in. In every scenario where anticipatory design made a choice on their behalf that they had NOT accepted in advance, it resulted in a negative experience. You should instead have a few hidden bonus services ready, that only appear to those users that the data show might benefit from them.

It is also worth noting that truly great anticipatory design requires a lot of data to truly understand a user’s needs and motivations. That is not to say that you should completely avoid it, but be aware that most users are still wary about giving away that much information about themselves when they can’t see the purpose of sharing them.

You need tons of Users or tons of Metadata

HBO underperforms when it comes to the users’ experience of personalization. Netflix is pretty good (they say their personalization made them $1 billion in revenue), but according to the users, it is actually still far from perfect, and YouTube nails it even better. So why is that? Well, Netflix has more meta-data than HBO, because they basically track everything and even let the user data generate their shows. They’ve been collecting data since they sent out VHS’s and DVD’s via snail-mail, and are by no means new to data-gathering. This, plus their huge amount of users gives them the edge over HBO but doesn’t explain how Youtube beats them at their own game.

Well, this is actually quite interesting, because YouTube has terrible metadata on their content compared to Netflix. Each uploader has their own way of describing and labeling the content if they even care enough to write anything. So instead of relying on this, YouTube uses a method called Nearest Neighbor. It finds users whose viewing history is most in sync with yours and starts recommending the videos in that user’s history that you haven’t watched yourself.

This method, mixed together with trending topics, channels you follow, and an extreme amount of active users, makes for a very effective personalized experience. Long story short, you either need to excessively label your content like Netflix, or have enough user engagements to implement a nearest neighbor based algorithm.

Have these Parameters in mind

To make a great experience, the users need to trust that your algorithm actually knows you well enough to make informed recommendations. If that is not the case, users will just ignore the recommendations as they fear it is going to be a waste of their time. There are two parameters that will influence a users’ trust in the algorithm: transparency and control.

Control

Allowing your users to express their feelings about content will make them aware that the service knows what you prefer. Most services allow you to like or dislike, rate, or otherwise inform the service of what you want more or less of. Some users want as little of this as possible, they expect the service to “just work”, while others enjoy expressing why they dislike that particular content (for example unsubscribe forms in emails allow the users to express why they don’t like that particular content).

Too much control can be overwhelming though, so strike a balance. In theory, the number of available feedback options could even be personalized individually, based on how much and how often each user interacts with it. A user on Netflix that often dislikes, and sorts through content, could get an expanded field, with the option to tick off why he/she dislikes it.

Transparency

Allowing users to inspect what behavior triggered a certain recommendation or UI change can be a very powerful trust generator. It can especially restore some faith in the service when a user feels that a certain recommendation or change doesn’t match their preferences, as long as an equal amount of control is provided. This way the user can either approve or disprove of each change, allowing the system to continuously improve. Take a moment to think of the last time you were frustrated with a service. Did they provide you with a way to adjust your experience for future use? Would it have made you more comfortable to continue using their service, knowing that this issue would probably be improved the next time? Of course, transparency and especially control also demands that you can actually follow up on the requests and feedback from the users. There would be nothing more annoying than telling a service that you don’t want to see x, only to find that x is still there the next time you log in.

On one hand, you don’t want your personalization engine to be a black box, but at the same time, some users don’t want to be reminded how much data the service has on them. I would recommend not putting that kind of detailed information up front, as it most of the time isn’t relevant for the user to look at, as long as the recommendations make sense. It is only when the service offers a recommendation that makes no sense to the user, that they will feel the need to inspect how it reached this conclusion and correct whatever mistakes it made. Make sure that when users are looking for it, they will find it, but they are not forced to look at it in their daily use of the service. The way you frame the data can have a huge impact too. Netflix used to have a category named “Because you like x, we think you will like y”. Some users didn’t like to be reminded that Netflix keeps track of everything they do, so instead they changed it to “other users who liked x, also liked y”. This actually made a difference to many of the users, who no longer felt like they were personally being tracked, and didn’t mind that other users were being tracked.

Trust

To make the users trust you, you need to find a balance. If your recommendations are too precise, and
the user has no idea how you get your data, it can get creepy, but if the recommendations are vague and often wrong, they might as well not be there (they actually have to be very precise for the user to even register that personalization is going on though).

This is what the aspect of trust is about. A user needs to trust that your recommendations are worth checking out, but at the same time feel comfortable that your service isn’t spying on their private life and collecting data it shouldn’t. Only then will your users continuously take advantage of your recommendations and this is essential if you want your algorithm to improve over time.

How do you get Users to Trust your Recommendations? was originally published in Adapt Tech blog on Medium, where people are continuing the conversation by highlighting and responding to this story.