Three principles for designing user-centered AI applications.
The significance of User Experience (UX) in the Development of AI.
The term Artificial Intelligence (AI) transcends the realm of computer science; it has always eluded a strict classification within this discipline. This is largely due to its intrinsic connection with a concept as natural as human intelligence. Each person holds their own interpretation of ‘intelligence’ — a quality that distinguishes us from animals, an attribute capable of evolution throughout one’s lifetime. Moreover, intelligence bears a mystical and indescribable quality, setting it apart from the more tangible concept of knowledge.
The term “intelligence” itself is so loosely defined that it sparks the imagination to overflow with possibilities.
The current state of AI research is shaped not only by computer experts but also propelled by the collective imagination of our society. The perceptions of AI held by amateurs significantly impact the trajectory of AI technology research.
Hence, in the design of AI applications, our responsibility extends beyond implementing cutting-edge AI technologies; we must also align with user expectations. Adopting a user-centered AI approach involves considering the experience users anticipate when interacting with artificial intelligence. The following three fundamental principles must be considered in the UX design for AI applications.
AI as acting!
The guaranteed path to creating a lackluster AI application is to develop a straightforward one that merely displays a static message like “This text is AI-generated from a month ago…”. Users expect more; they desire interaction with AI. Most successful applications operate on a request-and-response basis. Consider ChatGPT, for instance: it not only furnishes answers but also employs a typing animation, creating the impression that the AI is formulating responses in real-time. GitHub Copilot follows a similar strategy; it doesn’t just prompt me with suggestions when I log in and say, “Hey buddy, I read the code you typed yesterday.” Instead, it seamlessly integrates as an active element while the user is engaged in interaction.
AI as thinking!
We anticipate AI to contemplate the tasks assigned to it. Thinking, a pivotal facet of intelligence, manifests itself through a discernible delay in response. Just as humans pause to ponder and formulate the right answer, we expect a similar cognitive process from AI. ChatGPT doesn’t offer immediate responses; instead, it undergoes a processing phase, accentuated by a loading animation that conveys the impression of the machine engaging in thoughtful consideration. Consequently, this suggests that rather than instantly delivering our AI-generated results, introducing a slight delay — mere microseconds — can enhance the user experience, offering the sense of an intelligent machine without disrupting workflow.
AI as part of!
When discussing AI in software applications, we perceive it as an integral component rather than the entire system. A website, for instance, is not labeled as AI itself but rather acknowledged for having AI support or inclusion. Consequently, it becomes imperative to segregate the AI segment of an application from the rest. For instance, Copilot employs distinct color schemes, clearly demarcating the AI component from the rest of the application. This visual distinction enables users to swiftly identify AI-generated elements and information, allowing them to autonomously decide whether to agree or disagree — a critical option that upholds user autonomy.
I strongly believe that to achieve success, we must approach the incorporation of AI in our applications with a user-centric perspective, which I would term “User-Centered AI.” In contemporary software development, the emphasis on User Experience (UX) is rightfully significant. It is crucial to extend this focus to the integration of AI and establish patterns for the user-friendly utilization of AI. The three principles mentioned provide a guideline in this endeavor.
