PRODUCT DESIGN
What you need to know when designing products using AI and Machine Learning
In design communities right now, there’s been a lot of discussion around the emergence of Artificial Intelligence, the many possibilities it brings, and how to go about incorporating it into your design process. At this point, you’re probably asking yourself, “what exactly does it mean to work with AI as a designer? How do I design for Artificial Intelligence?”. You’re not alone in this.
When I started working on AI products a few years ago, I had a lot of questions myself, which is why I’m keen to share some of the key principles that have guided me and my teams over the last few years. With these tips under your belt, you can start to develop the experimental mindset you need to work on teams building products with AI and speed up value creation for your users.
Here are seven principles you should consider when designing a product using AI:
When designing a product, you shouldn’t rely on AI or Machine-learning to figure out what problems to solve. Instead, it is important to start with clear and specific goals that align with the business and user intent(s) for your solution. This will provide your team and product with a stronger sense of purpose and direction.
For example, in 2018, Google Docs implemented a machine-learning model to suggest grammar corrections in real-time as users type. While the model was successful in checking grammar accurately, users quickly realized that it was making suggestions that did not fit their writing style or voice. Millions of dollars later, Google learned that the accuracy of grammar was not as significant as providing suggestions that align with what the user may be trying to say. In this case, the lesson is simple: AI can provide powerful ways to solve complex problems, but its success ultimately depends on how well it aligns with the specific goals and needs of your product and its users. This is why you need to define your product goals based on user insights and then ask the question: How can AI help us solve this problem?
It’s essential to think about the various design decisions that will shape your algorithm when designing AI products. Consider the example of Facebook’s News Feed.
The News Feed uses machine-learning algorithms to show users the content that is most relevant to them. However, someone decided that it should be scrollable, that users should be able to react to posts with emojis, and that the content should be displayed in a particular way. These are examples of product and design decisions that do not rely on ML. In other words, while the News Feed is a product that uses ML, there are also several decisions that were made that contributed just as much to its success. By carefully considering these decisions, you can shape your algorithm to fit the needs of your product, rather than forcing your product to fit the algorithm. This will result in a product that is more efficient, effective, and user-friendly.
Designing with AI requires a continuous feedback loop. You need to be able to collect feedback from your users and incorporate that feedback into your product. This helps you improve the product and continue to provide value to its users. To create a feedback loop, you can design experiences that serve as an incentive for your users to teach the algorithm.
Facebook’s News Feed is another good example of this; by allowing users to react to posts, they are able to get explicit signals that represent what users want to see and refine their interests based on this. You can also implement various feedback mechanisms, like surveys, user testing, and analytics tools. By prioritizing feedback, you create a product that is more responsive to the evolving needs of your users.
It’s also important to think about how your product’s interface should adapt to user feedback and provide value in any situation. In some cases, every new piece of information provided by your users should affect how fast your AI product presents new results. For example, consider Google Maps’ location-based notifications.
Every action you take; like going to a new neighbourhood, or searching along routes brings you new information. This is why you should understand the ‘speed’ that is required of your product. By designing for adaptation, you can create a product that is more resilient, flexible, and responsive to the ever-changing needs of your users. In AI terms, this can be achieved through various techniques such as machine learning, dynamic algorithms, contextual awareness, situational analysis, and personalized recommendations.
While AI can automate many decision-making processes, it is not a replacement for human judgment and expertise. When designing user flows for your AI product, it’s important to consider how humans will be involved in the decision-making process. You should leave an open door for users to refine aspects of the AI’s output and provide help when needed.
Amazon, for example, allows users to remove items that might negatively influence their recommendations. Imagine you purchase a toilet seat on Amazon and you don’t need to buy another one for a while. Wouldn’t it be annoying to keep seeing toilet seat deals in your future personalized recommendations? Exactly. Hence why you should prioritise designing opportunities for your product to take corrections from its users. You can further achieve this through various techniques, like human-in-the-loop systems, decision support tools, and expert systems.
Context is critical when designing for AI. You need to consider the context in which your product will be used and ensure that it remains relevant to your users in that context. For example, in a search experience, the perfect recommendation is simply not enough.
Consider that the same word might have different meanings in different contexts. A search engine like Google makes use of machine learning to tailor your results to your location, search history, and time of day. So ideally, the answers your algorithm provides should always be contextualized in order to be helpful. And your role as a designer is to enhance this by making your user interface context-aware. By designing for context and relevancy, you can create a product that provides value to your users in any situation.
You’re excited to keep learning about AI x Product Design, aren’t you? Join my next class and pick up right where we left off. I’ll be sharing some more practical tips that have led to successful innovations for me and my teams. So come say hi and let’s dive into the world of AI design together!
