Creating user value with AI
6 themes for user-centred AI design
With artificial intelligence (AI) being the defining transformation technology of our time, as user experience designers we are finding ourselves having to navigate a tech-lead trend.
Businesses are rushing to implement and develop AI solutions and products to compete in a rapidly growing market. In 2016 it was estimated that AI will be a 70 billion dollar industry by 2020 and businesses are still just scratching the surface of leveraging the full potential of the technology.
However, this mission for us designers may sound too tech-driven. The AI and machine learning field still feels strongly dominated by computer scientists and developers and the voice of users often seems faint in the field.
But our oath as designers is to advocate for our users and make sure to focus on their needs. So how do we stay true to our calling and still navigate the business pressure to use AI in the experiences we’re creating?
The truth is AI is a fantastic tool. It’s actually so fantastic and powerful that it has the potential to solve enormous human problems, such as hunger and poverty.
The challenge in my opinion is not finding ways to use AI, because there’s an abundance of ideas for that. The challenge is to stay aligned to human needs and — as google’s people + AI guidebook describes it — “ find the intersection of user needs & AI strengths.
In an attempt to guide AI technology conversations to be more user-centred, I summarized 6 themes that can help ideate AI opportunities around user needs.
These can be a starting point for brainstorming and can provide your team with an anchor in user-centred goals for AI-powered systems or features.
The following is a detailed explanation of each.
1. Help me do my task more efficiently
This might be the first thing that comes to mind when thinking about how AI can help your users. If the user has to do a set of steps or actions more than once you can right away think of opportunities for automation. Think of auto-filling forms or smart defaults as ways to reduce the time it takes users to complete a task.
The system could also leverage historical actions to suggest the next actions or steps to users. Personalization opportunities can also be leveraged to surface likely next steps or frequently used items to help users find what they need to do faster.
AI can also be used to offer new ways to let users get to their goals. For example with the ViSenze online shopping plugin users can search the online store catalogues by image instead of using filters or searches that might take a longer time to understand and tweak.
2. Tell me something I don’t know
When I personally think of the one thing that machine learning or AI is good at, it is analyzing huge amounts of data very quickly. One way to think about the value of that is “faster than humanly possible?”. This can be leveraged to surface information, alerts or anomalies to the user. Things that they would not have been able to know because they cannot analyze data as quickly.
Some applications that address this are for example highlighting anomalies or relevant relationships in the users’ data. It can also help optimize inefficiencies in processes and detect patterns that could go unnoticed.
This can also be applied to tell the user what happened since the last time they were in the app or even in the world. AI can bring relevant information from news or other sources closer to the user, and highlight potential connections.
One of my favorite examples of this type of value is the German application Finanzguru. While the user tracks their expenses and regular contracts and payments in it, it can highlight potential savings based on local offers as way as payment patterns.
3. Become confident and prepared
Helping users be more prepared and predict potential issues is high up on the list of current applications of AI.
Whether it’s analyzing historical data and detecting patterns in them, the system can provide predictions and forecast trends that can help users be more prepared about what to expect in the future and of course, prepare the right action and make the right decisions accordingly.
The system can also become more proactive in making those recommendations and anticipating the risks before users are even thinking about this and would highlight potential risks or warnings early on. This can of course help users be more confident in their task in general and know that have more time to prepare and even prevent potential issues.
This can only work though if it’s coupled with providing users with the necessary transparency and explainability behind those predictions and recommendations.
Users need to understand the reasoning of the system as well as trace the data sources and be able to verify its objectivity and fairness in order to trust the system and follow its recommendations.
Research and design practices about AI explainability (XAI) and transparency are still in its youth but we can find good examples of how to support users to be more confident in their interactions with the AI system.
The above is an example from IBM Watson for Oncology where doctors looking at the recommended treatments can view derived attribute origins in the clinical note as well as view the evidence and rationale behind the selected treatment plans.
4. Help me learn faster and know more
There is an abundance of constantly growing knowledge everywhere around us, it’s almost impossible to stay up to date or retain all knowledge and information that is available.
Staying within the medical field, there’s for example continuous research, new discoveries, and experiments around the world. It would be impossible for a doctor to be at any given moment up to date with all the studies and research relevant to their cases. This is the core premise of Watson health, it allows doctors to be connected to an up to date to an ecosystem of health experts, data, and partners, saving their time and effort.
AI is also helping many researchers accelerate their process, though quickly extracting themes, topics, summaries, and mapping terms from documents, books, papers. These can be surfaced directly to users or can be made available on-demand through conversational interactions or chatbots. Allowing users to find the answers they need more quickly without having to read all the documents.
In general, helping users maintain the latest corpus and making sure they comply with the latest regulations is one of the most valuable things AI can offer.
AI can help relieve the user from the manual and time-consuming task and stress of this maintenance and can even bring up insights and information that users would not have gotten to due to the lack of overview.
A good example of that is the Watson application for HR block: Through referencing tax codes & laws, the system is able to suggest potential customized deductions & credits to the users to help users maximize their tax outcomes.
5. Support accessibility and inclusion
We’re just starting to scratch the surface of what AI can do in this domain. But think of all the visual, speech, and natural language recognition applications that can help users with diverse needs understand and interact with your application and with the world in a better.
A great example is Google’s live transcription app, which captions speech in real-time allowing people with hearing impairments to engage with conversations around them.
Providing users with an option for voice control instantly supports anyone with typing, touch or keyboard use limitations to make better use of your app, website or system. Conversational queries can go a long way with helping users who might find navigating through different pages and many clicks challenging.
Even a simple automatic language translation that can recognize the users’ preference or location and offers them the content in their own preferred language can make a lot of content more accessible to more users.
Furthermore, think of all the potential of leveraging IoT and smart sensors for making systems adapt better to users' needs, in common spaces but also in their homes. Think for example of additional features where users with accessibility needs can let themselves be known to the traffic light systems and request more time to cross.
Even just the simple awareness of the system of the users’ surroundings and environment can make many systems more inclusive. Sensors and devices that can help the system adapt to the users' light, sound or even habitual context would result in smoother and more accessible user experiences.
6. Delightful experiences
This is your wild card to unleash all the creativity.
Yes, addressing pain points and user needs is important, but what makes your users experience memorable is making it enjoyable and maybe even fun.
Referencing Aarron Walter’s hierarchy of user needs, it’s often the experiences that make sure to address all four types of needs and not just some of them that are the best and more compelling — also referred to as the pizza slice 🍕 by some researchers on my team.
The world around us is filled with examples of delightful and fun experiences that are powered by AI. Think of Snapchat which uses visual & facial recognition to detect face landmarks and apply lenses correctly. Or how Google Photos lets users search their photo library by emoji, a delightful touch to an already strong ai-powered search experience.
The above themes are high-level directions or intents that can help designers make sure the AI features they are building add real value to their users and address actual user needs. They can be used to ideate AI features around them for creating valuable and compelling user experiences. And you can find many ideas from designers about promoting people-first approaches in a technology lead world.
Of course, there are still more challenges with designing AI experiences that we need to understand and adapt our process to consider. But the key is to think of AI as a powerful capability that can be tailored towards use cases and user needs like other technology we’ve been designing with.
Laila Goubran is a user experience designer and AI design facilitator at IBM.
The above article is personal and does not necessarily represent IBM’s positions, strategies, or opinions.
References & further reading
