5 Principles for Designing Artificial Intelligence
One way for UX to take a more prominent role in the AI space is to advocate for users by creating heuristics by which to measure the value of intelligent systems for their intended users or purpose.
Applying the principles described below when designing AI systems, or evaluating existing systems according to these principles, will start shifting conversations about AI from a technical to a more user-centered perspective.
1. The system is being trained on diverse and representative data that reflects its intended audience
One of the promises of AI is non-biased / objective decisions, but this can only happen if our biases are not ingrained in the system. Designers / UX practitioners should use their expertise to help understand who is likely to be impacted by a system, and ensure that their perspectives are built into the system from the ground up.
2. The system reflects an acceptable level of “empathy”
While we’re still a long way (if ever) from an AI that can be truly empathic, we can ensure that we don’t let AI systems do what we wouldn’t let a human do. If the system is not treating users the way we would want to be treated (or would even be legally permissible), we need to rethink what the system is doing and how it is doing it.
3. The system gives users control
“User control and freedom” is a fundamental principle of UX design, and its importance doesn’t diminish because of the presence of AI. Experiences should be as personalized / machine-generated as users want them, not as we can make them.
4. The system has a clear intent and operates within those parameters, even if it can extend beyond them
Users should know when they’re interacting with an AI system, or when an AI is part of an overall system or service. Likewise, users should know, at a high level, how an AI system works, including the data it is using to make decisions. If users accept a system in one context, it should not be considered, by default, acceptable in other contexts.
5. Design still matters
The ideas that have emerged as part of UX (eg. information architecture, interaction design, usability evaluation, etc.) are still relevant. Data still needs to be structured in useful ways to train systems, and users still need to interact with these systems in the most intuitive ways.
AI is not a force obviating the need for good design. It needs to be viewed as a tool that can be integrated into already existing design frameworks to create compelling experiences. Using these principles to critically examine AI systems takes us a step closer, as designers, toward actively steering how AI gets incorporated into our daily lives.
