Product Design Principles in the Age of AI

Product Design Principles in the Age of AI

by Maggie McKosky

AI-powered technology is changing the way users interact with products, ultimately changing the way we build flows and interfaces leveraging this technology. With this shift, the principles of product design are evolving as well. To keep pace, UX & product designers can learn to adapt and apply new design principles to successfully incorporate AI technology into their solutions and products.

It's an exciting time to be a UX & product designer in this wave of AI technology. AI is prevalent in our everyday lives; we see it when we're online shopping, searching for items, and the platform then provides similar recommendations based on what we're looking at. AI also enhances the quality of content and imagery on digital platforms. We're even starting to see this with self-driving cars, which is redesigning the driving experience. These technologies today are brilliant at analyzing vast amounts of data to complete a particular task, a method used to do this is called Machine Learning, essentially improving its algorithms and performance over time as it gathers more data. While the possibilities of AI are endless, there's still room for improvement when it comes to transferring what it learned from one type of task to another, along with learning abstract concepts - the type of broad intelligence at which humans excel.

Even with this challenge, there's still hope. With a new approach to design thinking for AI, designers, engineers and product owners can bridge the gap between humanity and technology.

Design Thinking with AI

Design thinking, pioneered by international design and consulting firm IDEO, has three main pillars:

Empathy - Understanding the needs, motivations, problems of those you're designing for.
Ideation - Generating a lot of ideas. Brainstorming is one technique, but there are many others.
Experimentation - Testing those ideas with prototyping to validate ideas and assumptions.

When done right, design thinking captures the mindsets and needs of the people you're creating for, paints a picture of the opportunities based on the needs of these people, and leads to innovative and scalable new solutions through testing and iterating.

The same approach and design thinking framework can be applied when building AI innovations to teach machines. First, it's seeking to understand. In design thinking, it's about trying to understand the user, and in AI it's seeking to understand the data. This then leads to exploring ideas and potential models. Finally, it's prototyping and testing different solutions and models to deliver the desired results. In doing this, it's possible to create a scalable and repeatable process for building AI innovations. But there is one missing piece from the AI development process that is crucial - human empathy.

The AI development process typically starts with data and models instead of the user first; the user who is the benefactor of the outputs of those technologies. A new approach to design thinking is all about applying a two-pronged human-centered approach to AI development, resulting in five principles.

1. EQ in the IQ of AI - an Empathetic + Analytical Approach

The first principle is about finding and embedding emotional intelligence in the intelligence quotient (IQ) of AI; empathizing with your users using an analytical approach to understand their current issues, what's desired, and the gap in between. This starts with understanding the user, including behaviors and attributes, and how users navigate a tool or product while also understanding the data that is accessible. One way to help in this process, to better understand users, is creating a customer journey map, or persona if applicable. These techniques not only allow you to know the user better but also help identify pain points and opportunities, understand key decision points, and identify the data that might be predictors of key user decisions. Designing and building AI products will be less about delivering on the user's request and more about responding to the needs they haven't expressed yet. There's an opportunity to delight users through AI by providing them something they didn't even know they needed.

2. Define, Synthesize and Prepare

The next principle is defining, synthesizing and preparing what can be understood about users to come to a shortlist of the breakthrough questions and problems that can be solved through AI. Questions like: What capabilities does the user need at all times? What impedes the user's ability to complete their task? What is a hinder to user adoption?

Another way to help in this process of identifying your problem and uncovering data gaps is an opportunity analysis chart. This principle is a key step in diagnosing the user - understanding the needs, hurdles to consumption and satisfaction - and identifying the data needed to inform and drive the AI models being built. What information is your user giving you today? What do you know both explicitly and implicitly? What variables and data might be predictors of users' inputs and decisions that ultimately lead to better outputs?

3. Humans are the brains behind AI

The third principle implies that the brains in the new design thinking are really us; the end users who are using the outputs of the AI technology. Every facet of AI is fueled by human judgment; from the idea to develop a model and the sources of data chosen to train from, to the methods and labels used to describe it, all the way to the success criteria. The goal now is to generate ideas and solutions to the questions identified in the second principle. The way to do this is to come up with solutions molded from two streams of information. First, leverage the information about your user's problems and unmet or unknown needs. Second, pull insights from patterns and relationships buried in the data. Another tool to help in this process is an Opportunity Compass; an extension of the Opportunity Analysis chart in the last principle. It starts with the breakthrough questions you've identified previously and would like to answer, then list solutions you want to build and the levers you want to pull. Based on what models you decide to pursue, you can define the solution, the user's input and decisions you want to leverage, and the models you want to prototype and test. Then finally, it's about defining the three value propositions as to why you're building that set of solutions: 1) the user value, 2) the business value, 3) the learning model value. Why did you choose this model over the others? What specifically about the predictive results or outputs of that model are valuable? 

4. Testing Harmony

The fourth principle is all about testing. This is not a new concept, but it's not only about testing designs. Testing the front-end interface through wireframes and prototypes is one way to solicit qualitative feedback and validate solutions with users. Equally as important is to test the predictive outputs of your AI models to see if it truly delivers a solution to the problem you've identified. A crucial part of machine learning is the learning experience. Machine learning algorithms and computer vision technology need time and multiple iterations to learn; like flexing a muscle, it will get stronger over time. The first version of your model and its results will never be the best output; it needs to go through multiple iterations to gather more data, have more customers interact with it, and help teach the algorithm combined with human intelligence.

5. Continuous Improvement

The last principle revolves around the concept of iteration and continuous improvement in order to operationalize both the design and the AI models.

The key takeaway from the test, learn and release cycle is that any initial versions will never be the best iteration. Learning algorithms need more data, more customer inputs and outputs to learn, and therefore best understand right from wrong based on identified relationships and patterns in the data. The same goes for the design and overall user experience. The most important and hardest decision is trying to figure out how good is good enough for the final product.

The relationship between design and AI is constantly changing. Starting with your user first, not technology first, and asking some simple questions will lead you to success when designing using AI solutions. By integrating these five design thinking principles into the UX workflow and process, UX & product designers will have a better understanding of how to build AI in inclusive ways, considering both the technology and the user.

Maggie McKosky
VP of UX & Product Design at Shutterstock

Photo: Courtesy of Shutterstock