Discoverability of the Infinite
Discoverability is a design concept representing whether you can determine possible actions when presented with an interface or object.
As an easy example, below is an image of a clock radio designed by Dieter Rams. It has two simple controls on the front, "Volume" and "Tuning". Then, a few controls are on top to determine if you are setting the radio or the clock and a few other settings.
All in all, this is pretty straightforward to understand. If I gave you this radio, you could “discover” everything it can do in just a few minutes.
This radio clock was made in 1979, and technology has improved since then, with complexity increasing in step. This is the paradox of technology. As Don Norman puts it in The Design of Everyday Things, "The same technology that simplifies life by providing more functions in each device also complicates life by making the device harder to learn, harder to use."
Below are examples of iPhone screens over time. The original iPhone didn't have an App Store, so you can easily see everything the phone is capable of on one screen. Once there was an app for everything, we needed different ways to store our phone's new capabilities. The App Library. It is a place to keep all the apps we don't use enough to be on our home screen but still downloaded at some point. This creates a problem. How do we know it is there if we can no longer see an app on our home screen? We have to remember[1]. More capable technologies place an increased burden on our memory or become burdensome to navigate.
With the internet, we have infinite capabilities. More capabilities than we could ever hold in our heads. Below is an image of ChatGPT, a chatbot assistant. The interface is a simple text box with a few recommendations.
How can you help people discover the near-infinite capabilities of AI products? Once they discover the capability or new use case, how can they be assisted in remembering it? You could ask the model what it is capable of, but that is a long list, and it isn't even aware of many of its capabilities[2].
My best guess so far is contextual recommendations. Let the AI take in the world around you or other things you are doing on your computer, and then let it propose ways it may assist you[3]. It will require you to remember very little, but you will also surrender significant control of your life to an algorithm.
Ultimately, these products are unique because their programming does not limit their value. The value is limited by our ability to know what to ask for and determine what is worth asking.
I don’t have much novel insight here, but I think this space will be interesting to watch over the next few years to see how people tackle these design challenges. Maybe the interfaces will improve enough that I finally start using my Google Home for something more than listening to Spotify or checking the time and weather[4].
- A few apps I forgot I had until I dug through my App Library while writing this: Bing (they have an app?), IFTT, Pocket Points, Wavelength, Artifact, and Parsnip.
- These are called emergent capabilities.
- This has some negative privacy implications, so you wouldn't want to send all of that data back to a cloud. You would want to run the model locally.
- Previously we used the Google Home for "Animal of the Day" as well, but this was discontinued in November of 2023.
