It seems like we’re on the cusp of doing amazing things with chatbots and data mining, which can augment manual information architecture work and result in a better user experience overall. Automated information processing can help us identify search patterns and recommend information structures, to improve the findability of content.
How AI + UX = Better Findability
Let’s say you have a service that includes a search component. Right now, your users might be running searches and using manual filters to sort through the results, as search users are wont to do.
How do you know if they are finding what they need? You can rely on analytics to see if they are accessing what you want them to, and user research to find out if they consider themselves successful in completing their tasks. You can gauge overall user satisfaction by analyzing usage data, time on service, and ask them outright in interviews and feedback forms.
Let’s say some users have trouble searching or aren’t going where you expect from the search results. Maybe their overall experience is ok, but looking at the search terms and clicks through the search results, you can tell that there could be better results for their searches or clearer paths to the information related to their tasks.
How can you help them search better? Maybe if they used slightly more narrow terms or more comprehensive phrases, they might be able to find more relevant information to their queries. But these are theories.
You set to redesign the service, including the search, using a user-centered design approach. Awesome! And let’s throw in a little AI while we’re at it.
While you embark on your UX design process, you can use an AI system to analyze large amounts of seemingly unrelated data to help inform your design decisions. For example, you can set up your data mining tools to start collecting structured and unstructured data (analytics, search queries, and other usage data). As you identify which problem(s) you’re trying to solve for your users, you hook up an AI (like IBM Watson) to start analyzing the unstructured data.
But how does the AI system know what to do? This is the fun part: First, it parses the data at face value and then you have to train it. AI systems can analyze large amounts of data in much less time than could be done manually and can learn in real-time. They understand context so you can help them learn what the data represents by feeding them additional information in the form of business rules, metadata and questions.
As you work through the user experience research and design phases, you continually refine the questions you ask, and it will alter the data facets it analyzes. You can ask it plain language questions like: How many people search for X? How many times does Y get served as a response? What kind of information do we have about Z? The system responds to the questions as best it can, based on its analysis of the data. The beautiful part though is that you are not limited by your ability to ask questions. The system takes your questions and the data, and, actually learns. It starts to ask its own questions. Over time, as more queries are made in the search engine, and more user analytics are collected, it can better make connections, identify trends, suggest hypotheses, and generate richer findings.
How does this help users search? If your users rely on search to find information, you can augment the quality of the search results with this data. Think better predictive search terms, more relevant search results and Amazon-like cross-topic referrals. These have the potential to make for a richer user experience, as the content your users need is served directly to them by an engine that learns from everyone who came before.
AI for IA
And how can it help design better information architectures? One of the hardest parts of information architecture is creating appropriate content groupings with labels that are meaningful to users. Artificial intelligence can help discover and propose relationships between content, by analyzing content-related data for trends: from the meaning of the words themselves to how users navigate within it or search for it to how they move through the site or app or service, and beyond. AI is capable of highlighting trends that us mere humans might not see on their own, which can become information facets or new content use cases.
What if you could combine user research with large-scale data analysis performed by your AI system to better identify relationships between content types, and improve content groupings and cross-linking? To group content and label it in a more meaningful way for your users, to offer the right related links at the right time, and to generally make your site, service or product feel more intuitive. And what if it could analyze internal and external data to help you determine how best to build both internal information structures for content managers (e.g. for your content management system) and navigation structures for end users (e.g. the menu for your site or app)?
We still need humans
Now, to be sure, I’m not proposing machine-generated IAs (yet 😉 but I am suggesting that having an AI analyze user data from a number of seemingly unrelated sources can generate trends and relationships that might otherwise remain unseen. And it could provide a valuable input to decisions on how to structure content to better suit users.
If there’s anything we could use more of in UX, I’d argue that it’s not necessarily data; it’s intelligence. AI can bring us the intelligence we are currently missing about our disparate sources of data. And data visualization can help internal audiences better understand the outputs from the AI, to help inform decision-making. All of this is nascent, and it means that there is a promising opportunity for data scientists to become critical supporters of the UX Design process.
AI to IA to IM
In addition to supporting information architecture design, AI presents some really fun opportunities for information management by increasing the potential for latent findability and recommendation. Consider: What if you never had to tag the content you upload into your corporate document management system ever again, because the AI system can infer the meaning and relationships between documents?
What if your internal document management system could proactively notify you when someone uploads a document about a topic you’re interested in? And what if it could figure out that the document is relevant to your interests even if any particular phrase doesn’t explicitly appear in the paper, but the AI system can analyze the unstructured content in the document and map it to similar content, which you have bookmarked? How wonderful would it be to log into the system and have suggested content served up which is relevant to you and can help you do your job?
Where do we start?
In my case, I’m looking into plugging Watson into the back-end of the systems I’m building to start collecting data. I’m using user research to help inform the business rules I’ll feed into the engine. My goal is to generate better search results and eventually set up a chatbot-based recommendation engine to help users figure out what they need and who they should get it from because, let’s face it, no one knows how large organizations work and where to direct their requests.
And if I can at least help users figure out where to direct their requests, I can save them time and my organization a lot of money. It’s a small change with the potential for a huge impact.
But the key is to get started: find some project and add an AI component to see what it can do. Start throwing rules and questions at it to see what you (and it!) can learn.
Try. Play. Hack. Get started.