Insights UX Designers Need to Know at the Arrival of AI（Part II）
-Three suggestions for designers
In the next two years, many technology companies will dive into the wave of AI. As a designer, have you fixed your eyes on your company's new annual goal? Is "artificial intelligence" in the column? Is it mentioned a few times? What is the priority? Are you ready for this?
Perhaps AI has already been in your products. (if you are not sure, be alert), you may be thinking about how to improve the product experience through AI. Here are three suggestions for designers who have been or will deal with artificial intelligence.
Suggestions 1: "artificial intelligence" is not smart enough, please get ready to deal with the aftermath.
When AlphaGo is invincible, when the voice robots in the video know everything, when Silicon Valley technology sweeps across your network, it is difficult to believe the fact that: "artificial intelligence" is only a baby in the vast majority of areas of wisdom- This is what designers should keep in mind.
Do you offer a convenient "non-intelligent" option in your prototype in case to help users solve the problem if the machine's learning results are wrong?
When doing AI product design, always keep in mind the worst results of machine learning. The“worst-case retreat solution" is as important as the design in the best condition, and it is even more important most of the time.
Once the users are disappointed, they are likely to give up this function, or even the whole product. This kind of loss is difficult to recoup. So, in fact, one of the more important principle is that if you don’t have sufficient confidence in "machine intelligence", just give up.
How to clearly convey the benefits of "intelligence" to users, and how to provide elegant solutions for errors that may occur at any time - this is a challenge for UI/UX designers.
Suggestions 2: reducing the user threshold and giving timely feedback
Providing custom content for users is a very common application scenario of AI technology. The basic users information collection is often the basis for custom content. This sounds simple but is not easy to do.
Quora requires new users to select at least 10 areas of interest after the first registration is complete. And, users will be asked to manually input their familiar fields. We all understand why Quora is doing this. The intention of this interactive process is very good, but you can see a lot of users complaining on Reddit and Twitter.
Why? Two reasons:
1)The threshold is too high
Mandatory task should be simplified to the best extent before the product is used. The process should be optimized in the prototyping phase. In Quora’ s case, selecting at least 10 interests (not skippable) and manually entering the familiar fields is a threshold too high for users to bear. New users cannot see any content before the registration is over,and they even haven’t figured out what this site is about(there is only a login box on Quora' s home page). Under such circumstance, any mandatory, high-cognitive-load task may cause the lose of users. Users are not obliged to answer these questions, and they are not sure what they are doing this for.
Can you think of some optimization ideas to reduce this threshold? For example, is it possible to guess what the user might be interested in on the basis of the first two choices that users have made?( such as "Internet" and "design") Or is it possible to gradually understand their interests by analyzing the user's browsing behavior?
2)The feedback is not on time
If the acquisition of the initial data is mandatory and heavy, and users cannot feel the benefits of providing data, then "training artificial intelligence" is like a "voluntary job" - this feeling as a consequence is what we need try to avoid in product design. Therefore, we have to give correct feedback to users’ behavior as timely as possible.on Facebook,for example, if you make a friend application to a new colleague, you will find that Facebook will immediately update the list of friends you may know.
In fact, this feedback mechanism design in AI-related product is the same as it in game design principle. What we want to build is a positive, rhythmic short cycle like"first mission" -> "timely feedback" -> "motivate the user to complete the next task". Recall the games that you was addicted to in your boyhood, was it so?
Suggestions 3: Do not rush, accumulate trust by small things
If a newly graduated designer says he has the ability to redesign a better Facebook, you must think this person is insane. The trust between people is established on the basis of time and cooperation. Machine is the same.
All the "smart assistant" products on the market are still a long way to go from the real "smart". At present, the most important thing for AI products is to establish "user’ s trust", which is important but we have to be patient. We should start from reporting the weather right, turning the music on the right, and set the alarm clock right- we have to start from these standardized small tasks, and slowly win the trust of users.
Big companies often lack patience(which in fact gives a lot of opportunities to small AI teams), but we have to praise the e-business giant Amazon. Sitting on millions of Echo users, they resist the temptation in face of the attractive big cake namely "smart shopping assistant". They start from the small piece of cake of "repeat purchase". High frequency, low error rate - this is the entrance that all current AI products should be trying to find.
Most people do not like the feeling of “being guided”. And we do not like to lose control. But if a person continually completes every little thing you let him/her do, the trust is born. When the user slowly get used to buying toothpaste, cat food, kitchen paper by just one word on the Internet, maybe soon, the user will trust the company to do takeaway, setting ticket, booking hotel, and even buying a car and a house.
Credit is hard to build and easy to be ruined. Therefore, products designers, do not rush, get started from accumulating more credit points.
You may also like: