Using human-centered design to create smart devices people can trust.
We’ve all run into situations where one of our smart devices just isn’t acting so smart – like when Google Maps wants to send you to work on a Sunday afternoon or Spotify thinks you should only be listening to Nickelback. You’ve probably found yourself wondering:
- Why did my smart device just do that?
- How did it get that information?
- How can I get my smart device to work better for ME?
The trouble with making our smart devices smarter lies in the inherent limitations of the AI interface. With algorithms and machine learning, an AI’s output is going to hinge completely on the quality and context of the inputs funneling into it.
When AI is reading the wrong data, lacks context or the ability to take feedback, and applies insufficient reasoning, you can end up with results that are chuckle-inducing or simply irritating and frustrating. However at the other end of the spectrum, especially when coupled with a lack of privacy controls, poor or irresponsible AI design can create and exacerbate serious issues reinforcing cultural biases, creating abusive bots as well as surveillance and security abuses.
What happens though if we create an HMI interface for humans and AI that gives the user the ability to train, tweak and evolve the AI to make what’s smart even smarter. A human-centered design approach built on heuristics, architecture, usability and standards should create an AI interface where:
- The user can choose what data is right to use.
- The user consents to his or her data being used.
- The user is able to add context to the data being read.
- The user knows and understands the AI’s capabilities and limitations.
- The user knows the reasons for the results he or she receives.
- The user is able to provide feedback into the process.
In the end, humans need smart devices they can trust, control and rely on and that help them do what they want to do as efficiently as possible. A human-centered design approach can foster truly smart AI interfaces that know the user, know what to do, stay up to date, anticipate and take action to manage themselves all the while giving the user control and protecting privacy.
Contribution from Shasank Nagavarapu, Senior Associate Human-Centered Design