In the 2011 movie Contagion, you get the taste of what it would like to live through the outbreak of a pandemic.
The film anticipates some of the behaviours you might expect from citizens across the world. Panic buying loo paper, queues outside the supermarket, politicians addressing the nation on TV.
One behaviour that the movie makers didn't anticipate was a doubling in worldwide demand to learn a foreign language.
Yes that's right. When faced with the threat of a deadly pathogen many of our instincts aren't just to horde food and keep our families safe. It's to learn how to say s'il te plait ne respire pas sur moi or ask puedo comprar una máscara, por favor? And I was one of them.
In March 2020 I downloaded Duolingo and started learning welsh or dysgu Cymraeg if you prefer.
Why Welsh? I had toyed with the idea of learning Chinese or picking up French again, but my family is welsh so I feel likely to use welsh more often than I would another foreign language. Who cares? I'll come back to motivation later.
Why is Duolingo so good?
Duolingo is probably the most famous example of a product that has used principles of game design to help it's users learn a language. It's been so successful that it's the most frequently used education app in the world. Source.
Having spent a decent chunk of time reading about game design, it is a really fun product to use to try and spot what psychological techniques they're using to try and keep you coming back day after day.
Motivations Aligned
One thing to say about Duolingo is it's really easy to see how their user's motivation "to learn a language" aligns with what they deliver.
Duolingo's goal is to keep you using the app. If you're a free-user, keeping you on track means that they can show you more ads in between lessons. Or, if you're a Duolingo plus user, they want you to keep learning so you're less likely to unsubscribe.
In this way, it's fairly understandable that they use every psych trick, nudge and hack to keep you coming back - after all you want to learn that language don't you?
Juicy
The first thing you notice before you've even started a lesson is just how great he user interface is.
The colours, illustrations and animations make you feel like a kid in a candy shop. Even touching the regular button component gives a satisfying push (visual) and vibration (haptic) feedback.
The technical term that game designers strive for when designing a great interface is juicy.
Duolingo starts off on the right foot with a deliciously juicy UI.
Lessons
Like any language learning process, you take lessons. In Duolingo lessons are 10-20 questions long and focus on a specific topic - greetings, weather, animals etc.
They're bite size, and when I timed myself I typically find I complete them in around 90-120 seconds.
What I love about this is 90-120 seconds is the average time it takes to complete a race on Mariocart (150cc). So I wonder if there's a psychological trick here that humans seem to be able to focus comfortably for ~90 seconds and anything beyond that (e.g. Rainbow Road 3-5 mins) becomes tiring.
The next little thing I noticed was their use of sequencing to keep users in flow during a lesson.
Bare with me as I've brought some graphs to show you.
Now what lots of quizzes or exams might typically do is ease you in with the easy questions, then serve you the mid-weight questions, before serving the difficult ones at the end.
But from a UX perspective this is not very good. Because after a few easy questions you start to get bored, then you might be tested on some mid-weight ones, which take a bit of mental effort, before you're subjected to the genuinely difficult ones that take a lot of effort - by the end, you're mentally exhausted and feel like giving up.
This 'typical exam' sequence ignores the principle of 'Flow' that game designers try and keep their players in.
The 'flow' theory suggests that when playing a game or learning a skill, you want the user to be kept in flow. Flow is that wonderful state of mind where you're not too challenged as to be anxious, but not too easy so that it's boring. Flow focuses the mind in a pleasant way.
For Duolingo's designers, the task is to ensure that in each lesson players are being tested enough so that they're not too bored, but also not so difficult questions that they're haphazard guesses.
What they do really well is they'll often 'skill you up' in the first few questions, with verbs or nouns that are reappear in longer sentences further along in the lesson.
What I have noticed is they vary the difficulty of the questions throughout, with techniques like "tense and release" and "easing in and out" to ensure the user isn't too stretched or too comfortable.
From what I can see Duolingo normally finishes on an easier question so you don’t feel “frazzled” at the end of the lesson, which can build a sense of dread about starting the next one. The only time I seem to answer ‘hard’ questions at the end is if I’ve previously got one wrong and I’m trying to get it right on the second or third attempt.
Some UX designers use the concept of cognitive load and ego depletion, as if using your brain is like working out muscle in the gym. Just as if you were building up to doing some heavy exercise, your muscles get tired after a heavy load, so it is motivating to do a few lighter reps afterwards. A lot of the research on this is mixed, but I find it a useful metaphor nonetheless.
My guess is Duolingo have done huge amounts of testing to see how to keep users in flow based on their 'skill' levels. And I wouldn't be surprised if they have a machine learning model [edit: they do, called 'Bird Brain'] refining it for their millions of users every day. But the principles of flow come from intuitive game designers and are refined by the AI or algorithm.
Points, Badges, Leaderboards.
Whenever someone mentions gamification, points, badges and leaderboards are often three of the first 'levers' product teams pull to gamify their experience.
As I showed above, the basic unit of Duolingo is to take lessons. When you complete lessons you achieve experience points (XP).
In the free version, you lose "health" when you get a question wrong. Every day your health is refreshed so that if you run out you can come back again tomorrow.
When you complete a Level you get a crown. Crowns are good because they are shiny and royalty wear them. This is classic human psychology, we like to collect rare metals (gold) and we like high status (royalty) so crowns instinctively feel worth collecting.
The other 'item' games like to make us collect is fruit, because it's sweet and we like hoarding food. But Duolingo tends to go for shiny things like crowns and gems, where gems function as an in-game currency.
When it comes to badges, the way the Duolingo course structure is designed, makes you feel like you're earning badges by completing lessons and levels which in turn mean you can explore new topics.
Each topic has a pleasantly designed icon, which changes colour each time you increase in level. This, and their 'achievements' section is, In my eyes, how they're leveraging badges.
Each week you are placed in a league with 20 other Duolingo users around the world. Week 1 you start in the Bronze league and you are ranked on the leaderboard by your XP that week.
Here the obvious incentive is to get into the promotion zone, and avoid the relegation zone.
You can see how motivating it is by the speed with which people climb the leagues.
Speaking from experience, when you first get into Duolingo, this is really motivating. Because I am a saddo, this meant for a while my Sunday nights were often spent battling Eduardo who's learning Korean to get the top spot in the Emerald league.
These leagues are really motivating, especially for those who like to be seen as 'achievers'. And as you rise through the leagues each week you can start to feel like you're genuinely progressing your language skills, even though you're really just quite good at getting Duolingo XP.
Streaks
Above points and leaderboards, I think streaks are the most motivating tool in the Duolingo armoury.
There is a brilliant twitter thread about Streaks and how Duolingo coupled them with users' stated desire to learn a language and help them commit to the habit.
One problem they encountered with the introduction of streaks was people building up large streaks, breaking those streaks and then becoming really demotivated. I think what they noticed is a version of the "what the hell" effect.
The what-the-hell effect is a term coined by Dan Ariely, specifically in reference to habits around dieting. When people break their diet by eating a spoonful of ice cream, they might tend to eat the whole tub because "what the hell, I've already broken my diet"
As Ariely puts it
The what-the-hell effect is the feeling you get when you've already exceeded your preset limit and feel that since you've already failed, you might as well fail spectacularly.
In Duolingo's case, users had put in so much effort over consecutive days, and built up a 100+ day streak, once broken they would simply give up in despair and not open the app again because beating a large streak seemed so far away.
So Duolingo created a 'cheat' for users who'd broken their streaks could trade their "gems" to pay for a streak freeze, that would stop your streak from breaking and you can continue from where you left off.
As the above Twitter thread spells out, this was a really effective trick and one that's very popular with users.
I was one of these users, who used many a streak freeze to work my way to a 200 day streak. Woo!
But once I'd hit that milestone, something happened. Something unexpected. Something me of four months ago couldn't have anticipated.
I fell out of love with Duo.
The Divorce
As you can see above, for quite a long time I fawned over Duolingo.
I loved the juicy interface, I loved the lessons, I loved going up the leagues, I loved the notifications, I loved building up my streak but - over time -I'd stopped caring about learning welsh.
Huh?
Yes, I had stopped caring about learning welsh.
My motivation using Duolingo was to dysgu cymraeg. So I could converse with my welsh speaking family, so I could pronounce placenames more accurately, so I could sing the songs better.
But after nine months of using Duolingo, I just wasn't really that motivated to learn welsh anymore, and that felt sad.
Having reflected upon it for a while, there are two relevant insights from Game Design that help me understand what I went through. The first is a story about intrinsic vs extrinsic motivation.
There was a study that brought in some children who liked drawing. The experimenters left some pens and paper out for the children to play with in their break time.
In one third of study, the children were pulled aside and told they'd win a "good player" award if they did lots of drawings with the pen and paper.
The second third were just randomly awarded the "good player" reward only after they did some drawing. But they were unaware in advance that this was going to happen.
The third group didn't receive any rewards but were just given similar feedback on their drawings as the other children.
Two weeks later the experimenters left pen and paper out in the classroom again, and from behind a one way mirror, watched the children. Specifically observing how long for and how many drawings the same children created. Despite there being no rewards available this time.
They found that the kids who had been externally motivated with "good player" awards drew half as much as the kids who weren't previously rewarded by the experimenters for drawing.
What had happened, was by 'bribing' the kids with rewards, the external motivation (reward) crowded out the "intrinsic" motivation the children previously had for the love of drawing.
This felt very similar to what had happened to me. I had entered the process of learning welsh intrinsically motivated and through the process of adding extrinsic rewards - XP, Leaderboards, Streaks - I had forgotten what it felt like to want to learn welsh for it's own sake.
Designer Jesse Schell spells out the different kinds of motivation in his book The Art of Game Design.
Intrinsic Motivation: Because I feel like doing it for it's own sake
Extrinsic Motivation
- For payment (External)
- Because I said I would (Introjected)
- Because I think it's important (Identified)
- Because I'm that kind of person (Integrated)
The point that Schell makes isn't that extrinsic motivations are bad, they can be useful tool in any designers handbook. But they should be used carefully and we should be careful about any unintended consequences.
Schell again:
"They are both valid types of motivation and can actually work well in combination. However, sometimes the combination gets out of balance.
Free-to-play games often begin entirely focussed on pleasure seeking: big rewards, unexpected bonuses and exciting animations.
But over time they build up obligations - come back by a certain time or lose points, invite more friends or miss out on prizes.
Gradually, these games slide from pleasure seeking motivations to pain avoidance motivations. They keep you coming back, but you don't always feel as good about it.
For this reason, as designer Sheri Graner-Ray puts it;
People don't just stop playing these games, they divorce them."
This feels very relevant to the Duolingo team. Because most people I know have been through the same process of loving the owl and then coming to resent the tylluan.
Conclusion
Hopefully, after reading this you've learned a bit about how Duolingo lure you in with a juicy UI, keep you engaged with good lesson sequencing and motivate you with points, badges and leaderboards.
But hopefully you're also a bit clearer about the limitations of extrinsic motivations, and how it can come to crowd out the original motivations for picking up something new.
At the start I mentioned how Duolingo's incentives were very well aligned with their users.
Users came to start learning a language, Duolingo wanted to keep them there to show them ads or upgrade them to premium.
I'd say that's all true and fair up to the point where Duolingo's product starts to pull the user towards becoming better at Duolingo, rather than better at learning a language.
The cynic in me says Duolingo don't care. If users are coming back daily, or paying their subscription monthly, then they're doing a good job.
But if Duolingo is helping people start their journey on learning a language, but making them less likely to finish it, then in some sense what they're offering is not aligned with what their users want.
To some extent, if you want to learn a language, you're better off in the long term playing Words with Friends with your phone language set to Spanish than downloading Duolingo.
This doesn't make me angry with Duo the owl, just a little disappointed. I understand he's got bills to pay, that he was always well meaning and flawed character.
Through writing this blog post I have begun the process of forgiveness. And maybe, next pandemic I'll download Say Something In Welsh as a more sustainable way to enjoy the pleasures of dysgu cymraeg.
Diolch a nos da.
Addendum
I
An early reader of this post shared a video from Game Maker's Toolkit where the narrator explores this particular challenges that game-makers face. He goes one step further than me in suggesting some examples for how game makers can overcome the "divorce" affect.
Some of his terminology is different to mine:
E.g. where above I use the term sequencing, he uses the term pacing.
Where he uses the word engagement quite positively, I use the term engagement closer to what he means by addictive - as negative side effect of psychological techniques like variable rewards and extrinsic motivation.
One of the suggestions that caught my attention was how good games often encourage failure rather than make players ashamed by it.
To avoid any sense of obligation building up as your streaks rise, could a "freeze" day be reframed as a "rest" day and players be rewarded for it rather than taxed in gems? What affect would that have?
Likewise, instead of giving bonus XP for lessons where a user gets 100% of the questions correct, could you give more bonus points for a user who learns new words or phrases.
Another tip he gave was to add a sense of exploration to the game.
In Duolingo, the skill tree looks like a liner sequence with colourful lesson badges, how about they put that skill tree on top of a map of the country you're learning a language from?
I tried a quick mock up below, it sort of changes the vibe from "I am learning a language" to "I am on a quest to conquer Cymru!" Is this the right idea? I don't know but it would be fun to test.
II
This talk from a Duolingo engineering director shows how rigorous the A/B testing process is at Duolingo.
Most relevant to what I talk about above is her discussion of "measuring the immeasurable" (link is timestamped). In it she openly states it's hard to measure the impact of things like brand and design work, as well as long-term learning goals, which means in an A/B testing culture it's hard to prioritise those things unless you find a proxy-metric or defer to qualitative research.
She seems genuinely thoughtful about her students, so the cynicism in my conclusion is probably quite harsh, as she says she thinks about long-term learning outcomes every day - even though they're hard to measure.
I would love to know an example of something they've tested that has "worked" for their key metrics (DAU/revenue/app store ratings) but they chose not to implement because they sense it would harm the long-term learning outcomes.
It's easy to say play lip service to long-term learning but a principle isn't really a principle until it costs you money.
Send in any extra feedback you might have via the email link below. I really appreciate it, especially typos.
