UX & AI on the Same Team — Part II
In an article I published a few months ago, I discussed how artificial intelligence is impacting (and will continue to impact) the way in which we design. This is just one of the two major areas on which this technology has an impact. The other one is the improvement of the experience itself, that is to say, how to use artificial intelligence to improve people’s experience with our products or services, thus leading to enhanced satisfaction. I would like to reflect upon this latter aspect in this article.
Enhancing the experience
There is a lot of ground to cover when considering how artificial intelligence tools can contribute to more enriching experiences.
When designing, we must ask ourselves how we want the experience with our product or service to be and define the themes that will guide our design. With or without artificial intelligence, we must take into consideration several or all of the following dimensions when designing:
- Optimized experiences that anticipate people’s needs, reduce their workload, and simplify the tasks they must perform.
- Personalized experiences that adapt to each person.
- More human experiences that do not necessarily require direct interaction with fellow humans, but that aim for interactions with machines or people which feel more natural and create less friction.
- More transparent experiences that allow people to understand what is happening and that promote greater democratization in decision-making.
Before delving into each of these dimensions, I would like to make a disclaimer. What follows is an exercise in informed futurology, and I want to focus on the questions, rather than on the possible answers.
Optimized Experiences
When I speak of optimized experiences, I refer to understanding what a person wants to do and helping them do so with as little effort as possible. For quite some time now, several email platforms have introduced a feature that predicts text so that when we are writing an email, text suggestions to complete a sentence may sometimes appear on the screen. The system recognizes what we are writing, and it anticipates the end of a sentence, thus providing a prediction we can choose just by pressing on a keyboard key instead of writing several words. This is a great example of an optimized experience through artificial intelligence because:
- It understands what we are doing (in fact machines do not understand, they identify patterns).
- It predicts what we are going to do next.
- It reduces the effort to do so.
Now, what if we take this example to other situations? Is it possible to predict what transfers a person is going to make after receiving their salary based on their past behavior? Can we suggest, with little margin of error, what remedy a doctor will prescribe based on the results of a medical study? Can we understand if a student needs more math exercises than another one so as to learn addition?
In this regard, the questions that a design team must ask themselves are: How can we identify the task that a person is about to perform? And how can we make the completion of that task easier?
Personalized Experiences
Netflix’s home page is often mentioned as an example of ultra-personalization. Every person who uses the platform is supposed to have a different home page with recommendations based on their likes and preferences (and those of Netflix). And we must say that yes, it is a good example in terms of content personalization, but in terms of its layout, Netflix’s home page is the same for everyone.
By talking to colleagues who work in digital product design, I have come across the tendency to believe that artificial intelligence is a super-powerful back end that will serve personalized content within a design structure that is more or less the same for everyone. Indeed, that is one of the facets of personalization with artificial intelligence. But the less explored one is layout personalization, meaning the layout of an interface, the order of steps to perform a task, or the way in which we speak to users.
The questions that the design team must ask themselves at this point go beyond content personalization and they might be: Can we create different flows for people with more knowledge or experience than others? Can we adjust the messages so that they are more understandable and empathetic for each user? Is it worth it?
More Human Experiences
Suppose that a person has to give a system the order to move a file from one location to another. Grabbing that file and dragging it from one part of the interface to another one is much more human than writing down the instruction, as it used to happen in the old DOS before graphical interfaces. Now, that written order is a more human form of interaction than making a program on punched cards.
In that sense, I am inclined to think that one of the most natural interaction forms is conversation. Conversational interfaces, such as chatbots or voice assistants like Siri or Alexa, have been present in the market for several years now. To be honest, I do not consider that they have been as efficient as we expected some years ago, and many times the user experience is, at the very least, frustrating.
However, we are at the point when that can be reversed. It is likely that, just as easily as we can ask artificial intelligence to compose a haiku about Messi becoming a world champion today, in the very near future, we will be able to ask AI to schedule an appointment with our doctor for a weekday after six p.m. Therefore, the design team should ask themselves: What are the most natural or naturalized interaction forms? Is it really necessary to fill in some fields, click with the mouse, or swipe the screen to perform this task? Can we send a voice message instead?
I suspect that artificial intelligence will open up a very broad field to think about more natural forms of interaction.
More Transparent Experiences
By this, I mean that people need to understand what is happening and systems need to promote greater democratization in decision-making. When interacting with any system, we make decisions, and those of us who work in the field of human behavior know that it is not the same to:
- Guide people (make informed decisions).
- Persuade people (give them real arguments to make decisions that, we believe, are for their benefit).
- Manipulate people (use partial information and take advantage of biases so that they make decisions for our benefit, but not necessarily for theirs).
Today’s world is too complex, and there is so much information and supply abundance that sometimes it is difficult to make a decision and be really satisfied with it (regarding this topic, I recommend The Paradox of Choice, both the book and the TED talk).
For example, let’s suppose that we have to choose accommodation in a city where we will spend three or four nights. That task usually consumes a lot of time, and it can even be stressful. Even with a defined budget, a chosen neighborhood, and some other filter, the task of looking at the different offers, analyzing them, comparing them, and choosing one is complex. If well-designed and trained, AI solutions have the potential to reduce that complexity. They can silence the noise and overabundance to provide us with relevant information that guides us through the decision-making process. They can even make a recommendation that persuades us, as long as it has our benefit as its primary objective. Of course, if not very noble interests enter the design of these types of solutions, there is a very high risk of manipulating people.
The AI Interaction Paradigm
From my perspective, we are on the threshold of a new interaction paradigm. Until now, people have been giving instructions to systems, but we have not informed them about our objectives. Designing interfaces so that people can achieve their goals is part of the design team’s job. Here is an extreme example: suppose I want a hyper-realistic picture of a panda on the deck of a boat looking at the reflection of the full moon on the sea. That image can easily be created with Photoshop or a similar program. I am going to give that program, through the keyboard and mouse, different instructions (crop the panda, reflect the moon, arrange the elements in space, etc.). At some point, I will decide that the work is finished, and I will export it. The system will have received many small instructions, probably hundreds of them, but it will not know the result I want to achieve or my goal.
With generative AI this situation changes. Now, I can tell the system what my goal is, and that establishes a new interaction paradigm. We can extrapolate what happens with the composition of an image to any interaction. Instead of entering a banking system, going to the transfers section, entering the amount, defining a recipient, checking that everything is correct, and finally making the transfer, I can simply point out my goal: “Transfer ten dollars to John.” Or even be less informative and say: “Pay John what I owe him, and while we are at it, pay electricity, gas, and rent.”
One Step Further
But we can still go one step further. If we enhance the paradigm of telling the system our goal with the four dimensions already mentioned, and design more optimized, personalized, human, and transparent experiences, we will be able to achieve a large-scale leap in the quality of our products and services.
In this scenario, it is possible to send an audio to a virtual assistant and say, “I want to attend the next design conference in the city of San Francisco,” and let the system buy us tickets and look for flights at the time we like best. The system knows that I prefer the window seat so that other people do not wake me up if they need to go to the bathroom and it informs me about the advantages and disadvantages of choosing a hotel or renting an apartment for tourists, and keeping my benefit in mind, it gives me a recommendation about what it considers best for this particular trip or recommends activities to do in the city beyond the event. Getting a little into the field of science fiction, it is possible that our system resembles, even if only a little, Jarvis from Iron Man.
When designing experiences, we always seek that people achieve their goals with the least effort and the greatest satisfaction possible. With artificial intelligence as an ally, we can make a dramatic leap in this regard. To do this, we must ensure that AI helps us optimize, personalize, humanize, and make the experiences we create more transparent.
The challenge lies in integrating AI not only as a technological novelty but as a tool free of biases and manipulation, thus ensuring that it is used to improve the lives of individuals and society as a whole.
