AI meets UX – AI-based software calls for new approaches in user experience design

From brain mapping to speech recognition, we benefit from the advantages of artificial intelligence (AI) and have done so for some time now. It is part of every aspect of our lives. The way we do business is being transformed across all industries and sectors. This also affects how we design products and systems.

As we enter a new era of machine learning, the products we use do much more than just follow our commands. They also carry out tasks on their own. This has an impact on our expectations and the way in which we react. The task of UX design is to create trust in AI-driven software and generate useful, easy-to-understand products that make people’s lives easier and more enjoyable.

AI is everywhere in our daily lives. Our smartphones, cars, smart homes and voice assistants offer up a number of useful benefits.

Now is time to focus on the design aspect and implement AI-driven UX based on anticipatory and speculative design.

Anticipatory design – knowing what users want before they do

Anticipatory design combines the Internet of Things, machine learning and UX design. It is a discipline that aims to create automated, predictive systems that carry out actions proactively. So much is known about user behaviour that their needs can be anticipated. The system can trigger appropriate actions, making it possible to create an individual user experience that offers practical benefits to users and reduces the number of things they have to think about. Every day we have to make thousands of decisions. This can sometimes overwhelm us. Anticipatory design can be used to create products and systems that are able to make decisions for the user, reducing mental strain and freeing up time for more important things.

What if, for instance, you didn’t have to worry about what food to buy? That’s because the AI has already prepared a shopping list based on the dishes you regularly make. Here’s another example to illustrate the point: the user’s workout regime is adapted based on how they are feeling that day.

We already see the benefits of this today, if you didn’t know. Here’s how: Netflix knows what we want to watch. Spotify knows what we want to listen to. And WhatsApp suggests automatic replies to messages.

Of course, there are also potential issues that have to be considered. The main ethical concerns surrounding anticipatory design relate to data security and privacy. Another contentious issue is the bubble it creates around the user. Algorithms can create a loop that is difficult to break out of. Users are kept from having new experiences outside the typical events and activities found in their user profiles. It is therefore important that users retain control, that transparency is maintained and that users have the opportunity to opt out of receiving auto-generated suggestions.

UX design principles for AI experiences

Proponents envision a utopian future with artificial intelligence, while opponents see AI through a more dystopian lens. Navigating the space between these two points of view, the aim of UX design is to focus on users and their needs – not to push AI technology for its own sake.

As with any other product, the familiar interaction principles based on the ISO standard 9241-110 for quality UX design apply. But when it comes to AI-driven software, special attention must be paid to these four principles.

AI-driven UX principle no. 1: Visually differentiate AI content

AI-generated content can be very useful. But AI algorithms do not perform as they should if they do not have enough data to learn from. For that reason, it is important to visually differentiate the AI-generated content in order to let the users decide whether they want to trust it.

AI-driven UX principle no. 2: Explain how machines learn

Artificial intelligence can seem like magic at first glance. Sometimes even experts have a hard time explaining how the machine-learning algorithm produces the results it generates. Having information on how the system works and what data is used can be extremely useful for users. This can be done on an abstract level as well. One example is the product recommendations we are shown when we shop online.

AI-driven UX principle no. 3: Shape expectations

Self-driving cars have reached a level of technological sophistication where they are able to navigate even in complex situations. Despite this, we need to give our full attention to the system and temper our expectations accordingly. If we expect too much in the belief, for example, that a car could drive by itself without us, this could lead to misguided confidence and poor decisions.

AI-driven UX principle no. 4: Manage extreme cases

AI may generate content and perform actions that no one had thought of before. They come totally unexpected and need to be given greater attention in UX design. While they may strike us as bizarre or even humorous in some cases; they can also be disturbing and unpleasant in extreme cases. Here are a few examples involving chatbots where the context was not correctly understood and the AI produced an inappropriate response.

Clear communication about the AI’s capabilities can help users understand why unexpected things may happen.

A quick look at the evaluation metrics of an AI shows how to find the right balance between precision and recall. With recall optimisation, the machine learning product uses all the correct answers that can be found. Even if some of the answers turn out to be incorrect, an AI that can identify cancer will find all patients with cancer in a data set of X-rays using this evaluation metric. That said, some healthy people may also appear in the results.

When optimised for precision, the AI will only output X-ray images of the patients with cancer, but it may miss some of them. That’s because the machine learning algorithm only filters the unambiguously correct answers while potentially overlooking some positive borderline cases.

UX designers can provide developers with information about user expectations and priorities who in turn can optimise the algorithms to minimise inaccurate outputs.

Speculative design – exploring possible future scenarios

People and their needs should remain the focus of development work. UX designers should also look far into the future and anticipate social consequences. This is where speculative design comes into play.

We can utilise AI to optimally shape the way we live in the future.

To avoid unwanted consequences or developments, we need to start asking questions like: what kind of future do we want and what should life with AI look like? What impact could it have on the health and welfare system, on the financial system or even in the realm of politics? Where can disruptions occur and what is the potential impact?

Speculative design is a method of analysing possible future scenarios in terms of what positive or negative outcomes can occur and what measures can be taken to counteract them. This involves examining the many different future scenarios, which vary in terms of the likelihood of them occurring from very likely to possible someday. After this step, artefacts are developed that make it possible to experience and understand the individual scenarios. In order to make the right decisions you need to consider the long-term changes taking place.

Conclusion

It is necessary to protect personal rights, guarantee data security and provide legal certainty. UX designers are responsible for shaping how AI interact with us in the future and making sure that it is the user who gives the command (‘Hey Siri, turn off the light. I want to sleep.’) and not the other way around (‘It’s time for bed. I’m turning the light off now.’).