Software services

In reading Joe Dolson’s recent piece on the intersection of AI and accessibility, I absolutely appreciated the skepticism that he has for AI in general as well as for the ways that many have been using it. In fact, I’m very skeptical of AI myself, despite my role at Microsoft as an accessibility innovation strategist who helps run the AI for Accessibility grant program. As with any tool, AI can be used in very constructive, inclusive, and accessible ways; and it can also be used in destructive, exclusive, and harmful ones. And there are a ton of uses somewhere in the mediocre middle as well.

I’d like you to consider this a “yes… and” piece to complement Joe’s post. I’m not trying to refute any of what he’s saying but rather provide some visibility to projects and opportunities where AI can make meaningful differences for people with disabilities. To be clear, I’m not saying that there aren’t real risks or pressing issues with AI that need to be addressed—there are, and we’ve needed to address them, like, yesterday—but I want to take a little time to talk about what’s possible in hopes that we’ll get there one day.

Alternative text

Joe’s piece spends a lot of time talking about computer-vision models generating alternative text. He highlights a ton of valid issues with the current state of things. And while computer-vision models continue to improve in the quality and richness of detail in their descriptions, their results aren’t great. As he rightly points out, the current state of image analysis is pretty poor—especially for certain image types—in large part because current AI systems examine images in isolation rather than within the contexts that they’re in (which is a consequence of having separate “foundation” models for text analysis and image analysis). Today’s models aren’t trained to distinguish between images that are contextually relevant (that should probably have descriptions) and those that are purely decorative (which might not need a description) either. Still, I still think there’s potential in this space.

As Joe mentions, human-in-the-loop authoring of alt text should absolutely be a thing. And if AI can pop in to offer a starting point for alt text—even if that starting point might be a prompt saying What is this BS? That’s not right at all… Let me try to offer a starting point—I think that’s a win.

Taking things a step further, if we can specifically train a model to analyze image usage in context, it could help us more quickly identify which images are likely to be decorative and which ones likely require a description. That will help reinforce which contexts call for image descriptions and it’ll improve authors’ efficiency toward making their pages more accessible.

While complex images—like graphs and charts—are challenging to describe in any sort of succinct way (even for humans), the image example shared in the GPT4 announcement points to an interesting opportunity as well. Let’s suppose that you came across a chart whose description was simply the title of the chart and the kind of visualization it was, such as: Pie chart comparing smartphone usage to feature phone usage among US households making under $30,000 a year. (That would be a pretty awful alt text for a chart since that would tend to leave many questions about the data unanswered, but then again, let’s suppose that that was the description that was in place.) If your browser knew that that image was a pie chart (because an onboard model concluded this), imagine a world where users could ask questions like these about the graphic:

• Do more people use smartphones or feature phones?
• How many more?
• Is there a group of people that don’t fall into either of these buckets?
• How many is that?

Setting aside the realities of large language model (LLM) hallucinations—where a model just makes up plausible-sounding “facts”—for a moment, the opportunity to learn more about images and data in this way could be revolutionary for blind and low-vision folks as well as for people with various forms of color blindness, cognitive disabilities, and so on. It could also be useful in educational contexts to help people who can see these charts, as is, to understand the data in the charts.

Taking things a step further: What if you could ask your browser to simplify a complex chart? What if you could ask it to isolate a single line on a line graph? What if you could ask your browser to transpose the colors of the different lines to work better for form of color blindness you have? What if you could ask it to swap colors for patterns? Given these tools’ chat-based interfaces and our existing ability to manipulate images in today’s AI tools, that seems like a possibility.

Now imagine a purpose-built model that could extract the information from that chart and convert it to another format. For example, perhaps it could turn that pie chart (or better yet, a series of pie charts) into more accessible (and useful) formats, like spreadsheets. That would be amazing!

Matching algorithms

Safiya Umoja Noble absolutely hit the nail on the head when she titled her book Algorithms of Oppression. While her book was focused on the ways that search engines reinforce racism, I think that it’s equally true that all computer models have the potential to amplify conflict, bias, and intolerance. Whether it’s Twitter always showing you the latest tweet from a bored billionaire, YouTube sending us into a Q-hole, or Instagram warping our ideas of what natural bodies look like, we know that poorly authored and maintained algorithms are incredibly harmful. A lot of this stems from a lack of diversity among the people who shape and build them. When these platforms are built with inclusively baked in, however, there’s real potential for algorithm development to help people with disabilities.

Take Mentra, for example. They are an employment network for neurodivergent people. They use an algorithm to match job seekers with potential employers based on over 75 data points. On the job-seeker side of things, it considers each candidate’s strengths, their necessary and preferred workplace accommodations, environmental sensitivities, and so on. On the employer side, it considers each work environment, communication factors related to each job, and the like. As a company run by neurodivergent folks, Mentra made the decision to flip the script when it came to typical employment sites. They use their algorithm to propose available candidates to companies, who can then connect with job seekers that they are interested in; reducing the emotional and physical labor on the job-seeker side of things.

When more people with disabilities are involved in the creation of algorithms, that can reduce the chances that these algorithms will inflict harm on their communities. That’s why diverse teams are so important.

Imagine that a social media company’s recommendation engine was tuned to analyze who you’re following and if it was tuned to prioritize follow recommendations for people who talked about similar things but who were different in some key ways from your existing sphere of influence. For example, if you were to follow a bunch of nondisabled white male academics who talk about AI, it could suggest that you follow academics who are disabled or aren’t white or aren’t male who also talk about AI. If you took its recommendations, perhaps you’d get a more holistic and nuanced understanding of what’s happening in the AI field. These same systems should also use their understanding of biases about particular communities—including, for instance, the disability community—to make sure that they aren’t recommending any of their users follow accounts that perpetuate biases against (or, worse, spewing hate toward) those groups.

Other ways that AI can helps people with disabilities

If I weren’t trying to put this together between other tasks, I’m sure that I could go on and on, providing all kinds of examples of how AI could be used to help people with disabilities, but I’m going to make this last section into a bit of a lightning round. In no particular order:

• Voice preservation. You may have seen the VALL-E paper or Apple’s Global Accessibility Awareness Day announcement or you may be familiar with the voice-preservation offerings from Microsoft, Acapela, or others. It’s possible to train an AI model to replicate your voice, which can be a tremendous boon for people who have ALS (Lou Gehrig’s disease) or motor-neuron disease or other medical conditions that can lead to an inability to talk. This is, of course, the same tech that can also be used to create audio deepfakes, so it’s something that we need to approach responsibly, but the tech has truly transformative potential.
• Voice recognition. Researchers like those in the Speech Accessibility Project are paying people with disabilities for their help in collecting recordings of people with atypical speech. As I type, they are actively recruiting people with Parkinson’s and related conditions, and they have plans to expand this to other conditions as the project progresses. This research will result in more inclusive data sets that will let more people with disabilities use voice assistants, dictation software, and voice-response services as well as control their computers and other devices more easily, using only their voice.
• Text transformation. The current generation of LLMs is quite capable of adjusting existing text content without injecting hallucinations. This is hugely empowering for people with cognitive disabilities who may benefit from text summaries or simplified versions of text or even text that’s prepped for Bionic Reading.

The importance of diverse teams and data

We need to recognize that our differences matter. Our lived experiences are influenced by the intersections of the identities that we exist in. These lived experiences—with all their complexities (and joys and pain)—are valuable inputs to the software, serv

Picture this. You’ve joined a squad at your company that’s designing new product features with an emphasis on automation or AI. Or your company has just implemented a personalization engine. Either way, you’re designing with data. Now what? When it comes to designing for personalization, there are many cautionary tales, no overnight successes, and few guides for the perplexed. 

Between the fantasy of getting it right and the fear of it going wrong—like when we encounter “persofails” in the vein of a company repeatedly imploring everyday consumers to buy additional toilet seats—the personalization gap is real. It’s an especially confounding place to be a digital professional without a map, a compass, or a plan.

For those of you venturing into personalization, there’s no Lonely Planet and few tour guides because effective personalization is so specific to each organization’s talent, technology, and market position. 

But you can ensure that your team has packed its bags sensibly.

There’s a DIY formula to increase your chances for success. At minimum, you’ll defuse your boss’s irrational exuberance. Before the party you’ll need to effectively prepare.

We call it prepersonalization.

Behind the music

Consider Spotify’s DJ feature, which debuted this past year.

We’re used to seeing the polished final result of a personalization feature. Before the year-end award, the making-of backstory, or the behind-the-scenes victory lap, a personalized feature had to be conceived, budgeted, and prioritized. Before any personalization feature goes live in your product or service, it lives amid a backlog of worthy ideas for expressing customer experiences more dynamically.

So how do you know where to place your personalization bets? How do you design consistent interactions that won’t trip up users or—worse—breed mistrust? We’ve found that for many budgeted programs to justify their ongoing investments, they first needed one or more workshops to convene key stakeholders and internal customers of the technology. Make yours count.

​From Big Tech to fledgling startups, we’ve seen the same evolution up close with our clients. In our experiences with working on small and large personalization efforts, a program’s ultimate track record—and its ability to weather tough questions, work steadily toward shared answers, and organize its design and technology efforts—turns on how effectively these prepersonalization activities play out.

Time and again, we’ve seen effective workshops separate future success stories from unsuccessful efforts, saving countless time, resources, and collective well-being in the process.

A personalization practice involves a multiyear effort of testing and feature development. It’s not a switch-flip moment in your tech stack. It’s best managed as a backlog that often evolves through three steps: 

1. customer experience optimization (CXO, also known as A/B testing or experimentation)
2. always-on automations (whether rules-based or machine-generated)
3. mature features or standalone product development (such as Spotify’s DJ experience)

This is why we created our progressive personalization framework and why we’re field-testing an accompanying deck of cards: we believe that there’s a base grammar, a set of “nouns and verbs” that your organization can use to design experiences that are customized, personalized, or automated. You won’t need these cards. But we strongly recommend that you create something similar, whether that might be digital or physical.

Set your kitchen timer

How long does it take to cook up a prepersonalization workshop? The surrounding assessment activities that we recommend including can (and often do) span weeks. For the core workshop, we recommend aiming for two to three days. Here’s a summary of our broader approach along with details on the essential first-day activities.

The full arc of the wider workshop is threefold:

1. Kickstart: This sets the terms of engagement as you focus on the opportunity as well as the readiness and drive of your team and your leadership. .
2. Plan your work: This is the heart of the card-based workshop activities where you specify a plan of attack and the scope of work.
3. Work your plan: This phase is all about creating a competitive environment for team participants to individually pitch their own pilots that each contain a proof-of-concept project, its business case, and its operating model.

Give yourself at least a day, split into two large time blocks, to power through a concentrated version of those first two phases.

Kickstart: Whet your appetite

We call the first lesson the “landscape of connected experience.” It explores the personalization possibilities in your organization. A connected experience, in our parlance, is any UX requiring the orchestration of multiple systems of record on the backend. This could be a content-management system combined with a marketing-automation platform. It could be a digital-asset manager combined with a customer-data platform.

Spark conversation by naming consumer examples and business-to-business examples of connected experience interactions that you admire, find familiar, or even dislike. This should cover a representative range of personalization patterns, including automated app-based interactions (such as onboarding sequences or wizards), notifications, and recommenders. We have a catalog of these in the cards. Here’s a list of 142 different interactions to jog your thinking.

This is all about setting the table. What are the possible paths for the practice in your organization? If you want a broader view, here’s a long-form primer and a strategic framework.

Assess each example that you discuss for its complexity and the level of effort that you estimate that it would take for your team to deliver that feature (or something similar). In our cards, we divide connected experiences into five levels: functions, features, experiences, complete products, and portfolios. Size your own build here. This will help to focus the conversation on the merits of ongoing investment as well as the gap between what you deliver today and what you want to deliver in the future.

Next, have your team plot each idea on the following 2×2 grid, which lays out the four enduring arguments for a personalized experience. This is critical because it emphasizes how personalization can not only help your external customers but also affect your own ways of working. It’s also a reminder (which is why we used the word argument earlier) of the broader effort beyond these tactical interventions.

Each team member should vote on where they see your product or service putting its emphasis. Naturally, you can’t prioritize all of them. The intention here is to flesh out how different departments may view their own upsides to the effort, which can vary from one to the next. Documenting your desired outcomes lets you know how the team internally aligns across representatives from different departments or functional areas.

The third and final kickstart activity is about naming your personalization gap. Is your customer journey well documented? Will data and privacy compliance be too big of a challenge? Do you have content metadata needs that you have to address? (We’re pretty sure that you do: it’s just a matter of recognizing the relative size of that need and its remedy.) In our cards, we’ve noted a number of program risks, including common team dispositions. Our Detractor card, for example, lists six stakeholder behaviors that hinder progress.

Effectively collaborating and managing expectations is critical to your success. Consider the potential barriers to your future progress. Press the participants to name specific steps to overcome or mitigate those barriers in your organization. As studies have shown, personalization efforts face many common barriers.

At this point, you’ve hopefully discussed sample interactions, emphasized a

Ever since I was a boy, I’ve been fascinated with movies. I loved the characters and the excitement—but most of all the stories. I wanted to be an actor. And I believed that I’d get to do the things that Indiana Jones did and go on exciting adventures. I even dreamed up ideas for movies that my friends and I could make and star in. But they never went any further. I did, however, end up working in user experience (UX). Now, I realize that there’s an element of theater to UX—I hadn’t really considered it before, but user research is storytelling. And to get the most out of user research, you need to tell a good story where you bring stakeholders—the product team and decision makers—along and get them interested in learning more.

Think of your favorite movie. More than likely it follows a three-act structure that’s commonly seen in storytelling: the setup, the conflict, and the resolution. The first act shows what exists today, and it helps you get to know the characters and the challenges and problems that they face. Act two introduces the conflict, where the action is. Here, problems grow or get worse. And the third and final act is the resolution. This is where the issues are resolved and the characters learn and change. I believe that this structure is also a great way to think about user research, and I think that it can be especially helpful in explaining user research to others.

Use storytelling as a structure to do research

It’s sad to say, but many have come to see research as being expendable. If budgets or timelines are tight, research tends to be one of the first things to go. Instead of investing in research, some product managers rely on designers or—worse—their own opinion to make the “right” choices for users based on their experience or accepted best practices. That may get teams some of the way, but that approach can so easily miss out on solving users’ real problems. To remain user-centered, this is something we should avoid. User research elevates design. It keeps it on track, pointing to problems and opportunities. Being aware of the issues with your product and reacting to them can help you stay ahead of your competitors.

In the three-act structure, each act corresponds to a part of the process, and each part is critical to telling the whole story. Let’s look at the different acts and how they align with user research.

Act one: setup

The setup is all about understanding the background, and that’s where foundational research comes in. Foundational research (also called generative, discovery, or initial research) helps you understand users and identify their problems. You’re learning about what exists today, the challenges users have, and how the challenges affect them—just like in the movies. To do foundational research, you can conduct contextual inquiries or diary studies (or both!), which can help you start to identify problems as well as opportunities. It doesn’t need to be a huge investment in time or money.

Erika Hall writes about minimum viable ethnography, which can be as simple as spending 15 minutes with a user and asking them one thing: “‘Walk me through your day yesterday.’ That’s it. Present that one request. Shut up and listen to them for 15 minutes. Do your damndest to keep yourself and your interests out of it. Bam, you’re doing ethnography.” According to Hall, “[This] will probably prove quite illuminating. In the highly unlikely case that you didn’t learn anything new or useful, carry on with enhanced confidence in your direction.”  

This makes total sense to me. And I love that this makes user research so accessible. You don’t need to prepare a lot of documentation; you can just recruit participants and do it! This can yield a wealth of information about your users, and it’ll help you better understand them and what’s going on in their lives. That’s really what act one is all about: understanding where users are coming from. 

Jared Spool talks about the importance of foundational research and how it should form the bulk of your research. If you can draw from any additional user data that you can get your hands on, such as surveys or analytics, that can supplement what you’ve heard in the foundational studies or even point to areas that need further investigation. Together, all this data paints a clearer picture of the state of things and all its shortcomings. And that’s the beginning of a compelling story. It’s the point in the plot where you realize that the main characters—or the users in this case—are facing challenges that they need to overcome. Like in the movies, this is where you start to build empathy for the characters and root for them to succeed. And hopefully stakeholders are now doing the same. Their sympathy may be with their business, which could be losing money because users can’t complete certain tasks. Or maybe they do empathize with users’ struggles. Either way, act one is your initial hook to get the stakeholders interested and invested.

Once stakeholders begin to understand the value of foundational research, that can open doors to more opportunities that involve users in the decision-making process. And that can guide product teams toward being more user-centered. This benefits everyone—users, the product, and stakeholders. It’s like winning an Oscar in movie terms—it often leads to your product being well received and successful. And this can be an incentive for stakeholders to repeat this process with other products. Storytelling is the key to this process, and knowing how to tell a good story is the only way to get stakeholders to really care about doing more research. 

This brings us to act two, where you iteratively evaluate a design or concept to see whether it addresses the issues.

Act two: conflict

Act two is all about digging deeper into the problems that you identified in act one. This usually involves directional research, such as usability tests, where you assess a potential solution (such as a design) to see whether it addresses the issues that you found. The issues could include unmet needs or problems with a flow or process that’s tripping users up. Like act two in a movie, more issues will crop up along the way. It’s here that you learn more about the characters as they grow and develop through this act. 

Usability tests should typically include around five participants according to Jakob Nielsen, who found that that number of users can usually identify most of the problems: “As you add more and more users, you learn less and less because you will keep seeing the same things again and again… After the fifth user, you are wasting your time by observing the same findings repeatedly but not learning much new.” 

There are parallels with storytelling here too; if you try to tell a story with too many characters, the plot may get lost. Having fewer participants means that each user’s struggles will be more memorable and easier to relay to other stakeholders when talking about the research. This can help convey the issues that need to be addressed while also highlighting the value of doing the research in the first place.

Researchers have run usability tests in person for decades, but you can also conduct usability tests remotely using tools like Microsoft Teams, Zoom, or other teleconferencing software. This approach has become increasingly popular since the beginning of the pandemic, and it works well. You can think of in-person usability tests like going to a play and remote sessions as more like watching a movie. There are advantages and disadvantages to each. In-person usability research is a much richer experience. Stakeholders can experience the sessions with other stakeholders. You also get real-time reactions—including surprise, agreement, disagreement, and discussions about what they’re seeing. Much like going to a play, where audiences get to take in the stage, the costumes, the lighting, and the actors’ interactions, in-person research lets you see users up close, including their body language, how they interact with the moderator, and how the scene is set up.

If in-person usability testing is like watching a play—staged and controlled—then conducting usability testing in the field is like immersive theater where any two sessions might be very different from one another. You can take usability testing into the field by creating a replica of the space where users interact with the product and then conduct your research there. Or you can go out to meet users at their location to do your research. With either option, you get to see how things work in context, things come up that wouldn’t have in a lab environment—and conversion can shift in entirely different directions. As researchers, you have less control over how these sessions go, but this can sometimes help you understand users even better. Meeting users where they are can provide clues to the external forces that could be affecting how they use your product. In-person usability tests provide another level of detail that’s often missing from remote usability tests. 

That’s not to say that the “movies”—remote sessions—aren’t a good option. Remote sessions can reach a wider audience. They allow a lot more stakeholders to be involved in the research and to see what’s going on. And they open the doors to a much wider geographical pool of users. But with any remote session there is the potential of time wasted if participants can’t log in or get their microphone working. 

The benefit of usability testing, whether remote or in person, is that you get to see real users interact with the designs in real time, and you can ask them questions to understand their thought processes and grasp of the solution. This can help you not only identify problems but also glean why they’re problems in the first place. Furthermore, you can test hypotheses and gauge whether your thinking is correct. By the end of the sessions, you’ll have a much clearer picture of how usable the designs are and whether they work for their intended purposes. Act two is the heart of the story—where the excitement is—but there can be surprises too. This is equally true of usability tests.