Summary

Human-Centred Design (HCD) in AI refers to an approach and process where the primary focus is on the users’ needs, capabilities, and context. The design process begins with understanding the users, uses this understanding to drive development, and finally, validates the design against users’ needs in real-life contexts.

ELI5

Imagine you are designing a toy for your school friends. You wouldn’t just create anything; you’d find out what kind of toys your friends enjoy, what features they like, even what colors they prefer. Then you’d create a toy based on this information, and finally, give it to them and watch if they like playing with it. You’ll learn from their experience and make the toy even better. This is what Human-Centred Design in AI is like, but instead of toys, we’re making software or AI systems.

In-depth explanation

Human-Centred Design (HCD) is a process to design solutions deeply rooted in a comprehensive understanding of people for whom we’re designing. HCD in the context of AI encompasses not only a consideration of user interface and usability but also wider impacts, such as how an AI system will affect people’s daily life, the society at large and raise potential ethical considerations.

Three key stages of the HCD process are:

  1. Understanding users and their context: This includes extensive user research, needs assessment, ethnographic studies, and other methods. This stage aims to understand the users’ environment where the AI system will be deployed and provide insights into possible human-AI interaction.

  2. Developing design based on users’ needs and capabilities: Design ideas, prototypes, and iterations are created based on the understanding of users, their context, and needs. Prototypes can range from simple sketches to interactive mockups, wireframes, or early implementation of the AI models. Concentrated efforts are placed on ensuring that the system is not only functional and delivers a seamless user experience but also respects the users’ context and needs.

  3. Validating the design: The design is not static but continuously tested and modified based on real-world use and feedback. Validation involves user testing, field deployment, iterative refinements, and revalidation. This stage helps understand how the proposed solution works when users interact with it in a real-life environment and finesse it based on this feedback.

The HCD process helps to ensure AI systems are appropriate for the intended user and context, hence improve their effectiveness, efficiency, and user satisfaction. It also helps identify potential ethical issues early in the development process, such as invasions of privacy or issues of fairness, which can be mitigated or prevented.

User Experience (UX), Usability, Design Thinking, Ethical AI, Fairness in AI, Responsible AI, Privacy, Personas, Contextual Inquiry, Prototype Testing, User Research, Ethnography, Participatory Design