Late entry: I'm gonna start a @threadapalooza now on the intersection of behavioral science, product strategy, UX, and game design. I'm shooting for 100 tweets w/ as many individual opinions as I can before the end of the year. This is a product of a lot of thought over years
1. Every app is designed for behavior change, intentionally or unintentionally. People have agency, but their default behavior is swayed by their context. The design, functionality, etc. all influence user behavior, positively or negatively, predictably or unpredictably.
2. User behavior is a major input for the success of an app. If people use an app in suboptimal ways, then they are more likely to churn than people who are able to achieve their goals as a result of their behavior.
3. Suboptimal user behavior can lead to a product being undifferentiated from competitors. Imagine if you used Airtable but every cell was text format. Why not just use Google Sheets? 
4. Onboarding is absolutely crucial for building that initial mental model of how to use an app right and what's worthwhile to learn. Give a new user the confidence and information to be successful using the tools your app provides.
5. Ethics: We don't get to decide whether we influence user behavior or not. Flailing our arms blindly and ignoring obviously applicable behavioral science research is not a strategy.
6. The ethical decisions have to be made on a case-by-case basis. There is no simple rule for how to be ethical. However, things tend to go best when the user wants what the company wants, or at least there's some clear translation between the goals of both.
7. Having a business model with aligned incentives makes it way easier to influence people ethically. It's not impossible to ethically design for behavior change while having misaligned incentives (like is common with advertising business models), but it's much harder.
8. Example: Facebook's advertising business model means that they want users to pay attention to as many ads as possible, while users might prefer to just connect with family/friends through messages and comments. Facebook designs for attention, leaving users unhappy/controlled
9. Advertising business models generally want to increase frequency metrics like engagement. Increase the amount of daily active users, engagement duration per session, that sort of thing. Pay close attention to what users want. Does they want to increase their frequency metrics?
Alternatively, picture @quizlet. They are a flash card/study tool. They make money on advertising for free users and a paid subscription to remove ads and unlock new features. Many users want to increase their time studying, so that can ethically translate to frequency metrics.
11. In addition to making behavioral influence more ethical, aligned incentives also make behavioral influence easier and more effective. It’s amazing- people are more bought-in and open to suggestion from parties they trust.
12. Business models where the company depends on the user’s continued happiness for profitability tend to align incentives. 1-time purchases and subscriptions are pointing in the right direction.
We’re witnessing a zeitgeist in products playing w/ incentives. @SubstackInc is subscription vs ad/affiliate link supported writing, generating closer relationships b/w readers and writers. @Lemonade_Inc makes it so insurance isn’t hurt by claim payouts
14.Another great example of a biz model aligning incentives between end users and the company is @Superhuman, who charges a large amount of money, narrowing down their users into only the most invested power users. Superhuman mostly designs for power users firstround.com/review/how-sup…
15. I’m not going to dwell on ethics for too long, but it’s crucial to set the backbone for the behavioral design principles I’ll focus on. My approach to influence stems from a frame of being friendly and supportive to users, and much of it won’t work out of that context
16. A cool thing about cryptocurrency is that it lets you make incentives explicit and have total control over the rule systems for exchange. Incentives sort of form a base layer of behavioral influence. For example: medium.com/@virgilgr/ethe…
17. Communities run on cryptocurrency with smart contracts are a big opportunity, where everyone is invested and incentives are explicit. However, crypto adoption and (maybe) undermining intrinsic motivation is a challenge. Examples sourcecred.io fwb.help
18. The claim that a person's intrinsic motivation is undermined by extrinsic motivation is largely overrated and it's way more contextual/nuanced than people think.
Example: can we add an extrinsic motivator, like money or social status, for exercising? Someone who blanketly applies "intrinsic motivation is undermined by extrinsic motivation" might say no, it will backfire and remove all intrinsic motivation to exercise!
19. There often isn't any intrinsic motivation to undermine! In studies related to the subject, intrinsic motivation is operationalized as "they were already doing it without a reward." I'm not running at all yet, so an incentive could be helpful to get me started.
20. Extrinsic motivation can be useful as a way to get people started on a new behavior, and it can lead to developing new preferences that they didn't know they had. Perhaps I attempt to quit smoking for a money reward, but realize I feel better as a result!
(if it wasn’t obvious, the screenshots show citations/evidence)
21. Question: do meditators want to meditate as often as possible? A streak counter in @Headspace is an intervention for daily meditation. Many people appreciate the intervention because they want a daily consistency! Could users set their own goals?
https://twitter.com/RobertHaisfield/status/1343365902717177856?s=20
22. Extrinsic motivation can enhance intrinsic motivation if it engenders feelings of competence. If I want to establish an exercise habit but don't already have one, a reward marking the occasion of 10k steps in a day can make me feel capable of having an exercise habit.
23. Self-efficacy (the belief that you are capable of performing some action) is one of the most important factors out there to behavioral influence in product. Does someone believe they are capable of using your app successfully? If not, they'll probably drop off right away.
24. One of the lowest hanging fruit out there for increasing user retention in an app is to intentionally design what happens when users fail to reach their goal so that they try again rather than give up. robhaisfield.com/notes/intentio…
25. When people have high self-efficacy, they have a higher tolerance for friction and are more likely to adopt a flexible persistence with problems that arise. This means you can intentionally design for failure states by increasing upstream self-effiacacy.
26. People use apps in order to accomplish goals. They were probably trying to accomplish their goals elsewhere, unsuccessfully. The app can only help so much, user behavior needs to take them the rest of the way. No app is going to make me physically fit. I do that.
27. User failure is basically inevitable since a goal represents a discrepancy between where users are and want to be. You can't pretend it doesn't exist in your app. If you have an app that helps people diet, what is your plan for when people cheat and have a cheesecake?
28. A streak counter counts the number of days in a row that you do an activity. If you miss a day, then your streak resets to 0.
This is a poorly designed failure state that makes the user feel like their effort to build a habit was for nothing. Users often give up here.
This is a poorly designed failure state that makes the user feel like their effort to build a habit was for nothing. Users often give up here.
A streak counter could be improved by widening the partial failure margin and structuring streaks more like a health bar that goes down with each day missed rather than a "one hit kill" where you have to start over when you miss a day. mkremins.github.io/blog/failure-m…
@maxkreminski
@maxkreminski
30. A streak counter could be improved by giving the user a chance at redemption when they fail. Imagine if, upon losing a streak, you have a week to get 3 days in a row and regain everything you've lost? See my example of Hollow Knight here. robhaisfield.com/notes/intentio…
31. A streak counter essentially represents an avoidance goal, avoiding the negative outcome of losing your streak. People should think about what alternative interventions for frequency of usage could “count up,” making an approach goal to work towards.
https://twitter.com/ollybot_redux/status/1343659897154662409
32. Streak counters could be embedded within a meta-progression system. What if the design of an app could frame it as, "can you get a higher streak than last time?" It could show you your history and how you compare to your past performance.
(btw my point here isn't to endorse streak counters as a mechanic to build on but rather to show a clear example of a failure state and how failure can be intentionally redesigned)
33. Failure can be reframed. In Roguelike games, failure is the default state. Players expect to fail every time at some point! This means that failure is less memorable relative to success and players have the learning goal of improvement.
https://twitter.com/RobertHaisfield/status/1318623050867167233
An example of this in a non-game context is with coding. Writing incorrect code and fixing it is just what you do! Someone new to coding doesn't necessarily know that and may give up when they fail rather than debug. How could the design of an IDE reframe failure for new coders?
Aside- making failure less intimidating & giving users more approachable ways to respond to inevitable failure was a big component of designing the onboarding for GuidedTrack. Also, increasing self-efficacy by proving to users they could understand code
https://twitter.com/RobertHaisfield/status/1340371381750382592
When users successfully overcome failure, their self-efficacy goes up. In GuidedTrack, when people successfully fix an error, it means they (correctly) believe they'll be able to fix that and future errors again since they’ve done it already!
Users have to be able to troubleshoot for long-term retention. If the app is flexible and allows for a high skill ceiling as GuidedTrack does, then people will do more and more advanced things over time. This means they'll continue to experience failure!
https://twitter.com/RobertHaisfield/status/1343654412892311553?s=20
36 (last two counted). High user self-efficacy decreases a person's likelihood of dropping off (churn). Long term retention is influenced by successfully trying again after failure, which happens more when users have high self-efficacy.
https://twitter.com/RobertHaisfield/status/1343655543122976768?s=20
Imagine you're trying to establish a weekly review habit. It's Sunday one week, you've just done a lot of chores, & you feel so low energy that you don't want to do anything but watch TV. Instead of giving up and missing the day, you replace your hour review with a 10 min version
37Now you have a solution that works every time that "low energy" barrier comes up. In Spark Wave research, "habit reflection" was the most effective intervention for acquiring habits. Ask what has worked for you before, and apply to your present situation
https://twitter.com/RobertHaisfield/status/1265783898203553792
38. As a metaphor, I prefer "functionality" to "features." Functionality implies that it serves a purpose in the user's life. Features imply that it's something extra.
39. Onboarding tends to go better when people have a project. Learning the features of an app in the abstract is great, but only if that sparks their imagination about what sorts of projects to do. Projects place features into context, turning them into functionality.
40. Projects give users the motivation to learn an app, and show the user what they most need to learn. An app designed for both upfront and continuous onboarding
Example: Before I started work on GuidedTrack's onboarding, I designed and ran experiments to increase the likelihood of old people talking about death with their families. I learned exactly how GuidedTrack related to my goals. I want to generalize that experience for users!
41. When behavioral economists call people irrational, they’re more asking people to recognize where their behavior is misaligned with their goals. It’s instrumental rationality. Nobody behaves exactly how they want to behave, everybody needs help, this should be uncontroversial
42. People shouldn’t feel guilty or defensive about being irrational. If you recognize the gap between your intentions and reality, then that means you can take steps and accept help for behavior change.
43. When you successfully change your behavior for the better (improved sleep, exercise, information diet, etc.), it was because you made an effort to do something different from your default. Apps and tools can support and facilitate, but it only works if you overcome obstacles.
44. Tangential interlude: what would an open-source knowledge community maintaining a knowledge base of a domain of study look like? What other models exist besides Wikipedia's?
45. Other open question: how can you intentionally design for emergent, positive user behaviors?
46. Here I'll be talking about mental models a bit, so I'll defer the Naturalistic Decision Making model definition. sciencedirect.com/science/articl…
47. Epistemic rationality: has mental models and beliefs that are accurate reflections of reality.
Instrumental rationality: is able to effectively accomplish their goals.
In onboarding, I want to increase the user's epistemic rationality to enable instrumental rationality.
Instrumental rationality: is able to effectively accomplish their goals.
In onboarding, I want to increase the user's epistemic rationality to enable instrumental rationality.
48.A perfectly onboarded user is capable of operating the app to accomplish their goals, and can easily discover the solutions to novel problems they face over the lifetime of their software usage, or develop creative workarounds. They are instrumentally & epistemically rational
49.Intuition and mental models are built over time, as an expression of the tacit knowledge they gained through past experience. Intuition doesn't come on day one. Could you imagine expecting to learn how to do everything you'll ever want to do in Excel in the first ten minutes?
50.When I conceptualize onboarding, I split it up into upfront onboarding and continuous onboarding. Upfront onboarding is the learning that happens in the first 10 minutes. Continuous onboarding describes continuously learning how to use an app better over time.
51.Upfront onboarding sets the stage for users to develop intuition, but intuition is developed through experience over time. Users build their intuition during the continuous onboarding, in response to the feedback they are given on the impact of their behavior.
52. Feedback loops are the process with which we can update someone's mental models and intuition. Both game design and Perceptual Control Theory discuss this in the context of behavior, in mostly similar ways. Image is an amalgam en.wikipedia.org/wiki/Perceptua…
53.Based on the discrepancy b/w the app's state and it's desired state, users can:
-update their behavior
-update their desired/predicted state
-update their predictions of the outcomes of their behavior.
The loop repeats, eventually leaving the user with mastery over their tool
-update their behavior
-update their desired/predicted state
-update their predictions of the outcomes of their behavior.
The loop repeats, eventually leaving the user with mastery over their tool
54.Designing a feedback loop to influence user behavior is about showing users the discrepancy between current and predicted/desired states of reality, while also ensuring that their causal predictions of outcomes of their behavior on the system make are as accurate as possible.
55. The accuracy of a person's predictions reflects how closely their mental models reflect the software's actual feedback rules. In other words, it reflects how epistemically rational they are in the context of using an app well.
https://twitter.com/RobertHaisfield/status/1344770968934207489?s=20
56. As opposed to day-to-day life, product design is able to dictate the outcomes of user behavior and how that is made known to users (feedback and feedback rules).
You can design apps so people exercise their autonomy to gain competence and expertise.
You can design apps so people exercise their autonomy to gain competence and expertise.
https://twitter.com/roberthaisfield/status/1286279650973372417
57. Self-efficacy operates on feedback loops. When people are successful or overcome failure, their self-efficacy increases, leading them to try more things. People feel confident in their competence.
When they fail to convert failure into success, they feel less competent.
When they fail to convert failure into success, they feel less competent.
https://twitter.com/roberthaisfield/status/1343653565861007360
58/Feedback loops aren't always triggered. Do you notice your breath rate? The Spire Stone is a wearable that measures your breath and makes you aware of your rate through distinctive vibrations so you can slow down if it is too fast. The device makes the feedback loop salient!
59. "Intuitive design" is less about designing something that is intuitive and more about designing something that builds intuition through feedback loops. This distinction is especially important if you're building an innovative product, where prior intuition does not exist.
https://twitter.com/roberthaisfield/status/1344776652790796288
60. Two examples. In @RoamResearch, many of its users have never used an outliner before to structure knowledge. They definitely haven't used bidirectional links before. The fact that you structure your pages through their indentation position in a block hierarchy isn't intuitive
https://twitter.com/roberthaisfield/status/1344354222658109440
I've directly worked on this for GuidedTrack. It's built to be learnable, not intuitive. One of our user groups involves researchers looking for alternatives to Typeform and Google Forms. GuidedTrack's text interface will not be intuitive to people who are used to drag and drop.
62. Writing code in GuidedTrack is not hard. In fact, it's incredibly easy to grasp and we've got tools built in everywhere to help people get started. But many users have never programmed before and their mental models need to update from "scary" "for programmers" "drag & drop"
63. Before I started working with the onboarding, users had a long feedback loop of writing code in a text editor, clicking preview to see the outcome, and then making adjustments to code. It took a while for the user to go from reading code to understanding its output.
https://twitter.com/roberthaisfield/status/1340371373730885633
64. A live programming environment, where the user saw their code side-by-side with the program it produced, speeds up the rate of people realizing that they can, in fact, read the code. It allows people to adjust their code and immediately see the output.
https://twitter.com/roberthaisfield/status/1340371376079704065
65. Feedback loops are more efficient at communicating the workings of a system than words. We were literally able to remove code comments explaining what was going on because the user could just see what was going on with the new side-by-side preview/debugger
https://twitter.com/roberthaisfield/status/1340371381750382592
66. Live programming environments fall into a general category of “learnable programming.” While we originally came to the solution on our own, finding out that there’s intellectual history to it enabled us to improve our design. @worrydream worrydream.com/LearnableProgr…
67. All of my best ideas were stolen from me years before I came up with them 🙃
• • •
Missing some Tweet in this thread? You can try to
force a refresh
