Automagic Revisited

Last Xmas I wrote a piece about The Phenomenon of Automagic. I defined Automagic as when your users don't know how your app is working - it just works. Last week I was the OzCHI (The Australian Human Computer Interaction Conference) and Abi Sellen from Microsoft Research opening Keynote made me give automagic a second thought. 

Abi spoke of Symbiotic Design - designing systems that mutually shape our behaviour. About making our behaviour intelligible to machines, as well as making machine intelligible to humans so users can understand the interface and use it effectively. 

When automagic goes too far, it becomes Artificial Intelligence which scares people, as we can see from Google. 


Through Abi's research at Microsoft she had seen instances where people don't trust the automagic. Doctors and clinicians want to understand how a computer comes to a diagnosis, so that they can trust it. In another case, algorithms hidden to users led to frustration and misuse. When the algorithm was shown, through faster feedback, users shaped their behaviours in response to the system. This makes the design more obvious and helps learnability. 

Rather than creating systems that see, think and act as people do, focus on extending and augmenting people's capabilities - Abi Sellen

While I still think automagic has its place, it can co-exist with symbiotic design; too much automagic comes at a cost when it makes the interface hard to understand. Users need to understand what the computer is capable of:

  • Sensing and capturing
  • Perceiving
  • Recognising
  • Doing in response to input

The output and feedback from a system needs to be intelligible to users so they can use the system. As Abi says, we needs to stop designing intelligent machines, and start designing intelligible ones.