Every Label Is a Social Choice: What Sociology Taught Me About Designing for Mental Models

How we interpret things in a digital interface depends on our ideas about the world which comes from our cultural frame which in turn comes from our individual socialization. When we name a tab “Resources,” “Help,” or “Courses,” we’re not just labeling buttons — we’re saying how we think people see and understand the world. In UX, that set of expectations is a mental model — our best guess about how users might view and understand something. Sociologically, those models are learned over time through socialization — family, school, work, and the cultures we move through. As a Sociologist, I read design sociologically: mental models and labels aren’t just cognitive shortcuts, they’re social products that reflect whose experiences and norms are being centered and whose are ignored.

The same word can mean different things to different people, depending on who’s reading it. A 20-year-old student, a faculty member, and an international visitor might read “Community” three different ways. No one is “wrong”. They just think differently because in the real world their roles and contexts are different, which stems from their distinct socializations. Good labels don’t just name features; they help all user classes feel like they belong and don’t marginalize any user class.

In our project to re-architect the Canvas Index and AT@UT page, we began by defining our key user classes because each group brings a distinct mental model. Those mental models shape how people search, label, and prioritize tasks.

What Mental Models Are and Why They Matter

A mental model is simply a person’s understanding of a system. Take “Canvas Index” page. To UT staff, that sounds obvious: a list of course stuff. To a first-year student, it’s murky. An index of what — classes, modules, grades? In our card sorts and interviews, faculty grouped content by administrative function (“Assignments,” “Modules,” “Announcements”), while students grouped it by goal (“What I need to do,” “What I need to know”). It reflected two different mental models built by two different roles and contexts. Therefore our task as designers is to re-design the interface taking into account mental models of all our user classes and have label language that is akin to the language used by the different user classes without confusing and alienating any user classes.

How to Elicit Mental Models (and Turn Them into Labels)

The first job is to understand how each user class makes sense of the digital interface. Don’t chase the “right” wording in a vacuum — ask people how they already sort and name things, then map your labels to their language.

For the Canvas Index and AT@UT pages, we ran an card sort with 8 students and 3 faculty. It surfaced two distinct mental models: students grouped by goals (“what I need to do/know”), while faculty grouped by administrative tasks (build, publish, grade). In retrospect, we missed new students and faculty, whose first-time perspective would likely have pushed us toward more recognizable, action-first labels and people using assistive technologies, which would have ensured that the interface would be accessible to all. This omission matters: whoever you recruit ends up becoming the “default user” your interface serves. How to do it better next time:

  • List your user classes up front (e.g., new students, returning students, TAs, new faculty, returning faculty and students/faculty using assistive tech).

  • Sample intentionally, not conveniently. Aim for a small but diverse panel (e.g., 5 per user class) rather than one large, homogeneous group.

  • Mix methods. Pair card sorts (how people group/name) with 15-minute interviews (how they talk about doing things).

  • Map language to labels. Capture exact phrases users say and reflect them in labels.

  • Validate across classes. If a label works for faculty but confuses students, iterate until it’s legible to both.

  • Account for newcomers. Always include first-time users; their mental models reveal the “recognition vs. recall” traps you can fix early.

  • Account for users using assistive technologies. This makes sure that the interface is accessible to people with different abilities.

How People Scan and Read Labels

Most people scan interfaces — they don’t read every word. Scanning patterns (e.g., F-shape, quick skims) mean labels must be legible at a glance. To make labels scan-friendly and inclusive:

  • Lead with the action. Put the verb or task word first (e.g., “Submit Assignment,” “Merge Courses”).

  • Keep it short and concrete. Aim for 1–3 words when possible.

  • Avoid institutional jargon unless your users know and use it.

  • Support linguistic diversity. Pair concise labels with brief helper text or familiar icons to clarify meaning.

  • Mind directionality and accessibility. Account for left-to-right vs. right-to-left reading, screen readers, and low-vision users.

The goal isn’t minimalism for its own sake — it’s clarity that helps every user class find and complete tasks quickly.

Adjusting Labels to the Site and the Audience

Labels should match the goals of the users and the task the user intends to accomplish on your site. It’s important to remember that different user classes have different mental models. UX designers and researchers should keep this mind and take into consideration multiple mental models including the mental models of marginal user classes. For example, the UT canvas index and AT@UT pages have multiple user classes such as new and returning students, new and returning instructors, TAs and users with different abilities. The interface should be designed keeping in mind the mental models of all its user classes. Labels and navigation should reflect each user class’s top tasks:

  • Students: “Student Tutorial”

  • Instructors/TAs: “Merge Courses”, “Gradebook”, “Sandbox”

What I Learned Designing Labels for the First Time

Sociology pushed me to chase fit — fit with how people imagine the system. This means that I try to fit user’s mental model with the system’s conceptual model. One student called Canvas “my academic hub.” A faculty member called it “the course management system.” Both are valid. Our redesign reconciled these perspectives by:

  • Using role-based top-level categories that match intent: For Students | For Instructors | Resources for All

  • Writing sub-labels in users’ own words from card sorts and interviews (e.g., Merge Courses, Gradebook).

Practical Guidance to Effective Labeling

  • Recognize and define each user class. Recruit people from each user class; Do open card sorts; Follow with short interviews to capture real phrasing.

  • Lead labels with verbs.

  • Test with all major and marginal user classes (e.g., new and returning students, new and returning faculty and people with different abilities).

Final Reflection: Designing for Everyone

If we want information to be findable and understandable, we have to start with how people make meaning. Every click begins inside a cultural frame — learned over years of socialization — and every label encodes a mental model. Good information architecture leverages existing mental models to build “intuitive” interfaces. However, mental models aren’t universal. Designers inevitably privilege some and sideline others. So whose models get built into a digital space? Usually those of people with power and proximity to the design process: decision-makers, legacy “core” users and the segments that drive revenue. This is amplified by treating frequent, returning users and those with high digital literacy as the norm. In short, the “default user” is a social construction, not a neutral fact. This is where information architecture maps onto social architecture: labels mirror existing hierarchies about who counts, who is “typical,” and who is an “edge case.” Sociology helps name and analyze those hierarchies — showing how class, race, gender, disability, language, immigration status, and technical abilities shape whose mental models are centered and whose are ignored. Practically the solution to diagnose this problem means widening whose mental models inform the information architecture which means recruiting beyond convenience samples, co-designing with first-time, low-digital-literacy, multilingual, and assistive-tech users, vetting labels with the communities and treating so-called edge cases as primary use cases. The best label isn’t the clever one — it’s the one that fits your audience’s world and invites them to complete the task with confidence.

When Information Architecture Mirrors Social Architecture: Lessons from Craigslist

When we looked at Craigslist in my Information Architecture class and other students called its Information Architecture “good,” I was stunned because to me, it felt like cognitive overload. The way content is organized in a digital space often assumes certain know-how or mental models. To me, Craigslist looks like an artifact from the past: long blue lists, almost no visual hierarchy, category pages that feel like spreadsheets. To me, information architecture’s primary goals are to organize, structure, and label content so that users can: find information easily, navigate the system intuitively, understand their location and context within the information space, and complete their intended tasks with minimal frustration. To me, Craigslist helped with none of them. I can’t find information easily because of ambiguous categories, inconsistent labels, and insider jargon that assumes you already know where things go. Navigation doesn’t feel intuitive because the site leans on long, flat lists with little visual hierarchy and no clear pathways between related sections. I rarely know where I am in the site because there are no breadcrumbs, weak page titles, and minimal context. Completing tasks is harder than it needs to be.

And yet, in my class, I heard that as an information architecture, Craigslist undeniably works as millions of people still find apartments, furniture and gigs there all the time. Clearly, something is working, even though Craigslist looks like something from the 1990s. In class, I learnt that the very things which appeared to me as “bad” about its information architecture are also what make it surprisingly “good.” Understanding that tension reveals something bigger about how information architecture succeeds for people who already know how to navigate a system and how it can fail those who don’t.

What Craigslist gets right is its simplicity. Its structure has barely changed in decades: a set of top-level categories, consistent subcategories and predictable pages. That consistency builds a reliable mental model. If you’ve used Craigslist once, you can use it again without relearning where things live. There’s also very little visual noise. Because the pages are mostly text, the “information scent” is clear for experienced users: scan the category, skim the list, click decisively. This plainness also means fast load times and decent performance on weak connections, which matters for people on older devices or limited data plans or slower internet. In short, Craigslist lowers the cost for repeat users. For power users who have internalized Craigslist’s grammar, the site is legible, quick and efficient.

But those same traits turn into barriers for newcomers like myself. The flat, list-heavy structure depends on recall rather than recognition: you need to know the right category label before you can even start. Is that short-term room under “sublets/temporary,” “rooms & shares,” or “vacation rentals”? If you don’t already know the site’s grammar, you face friction immediately. This is where Craigslist’s information architecture reveals something about social life as well as design. What many designers call “intuitive” design often means “familiar to insiders.” Sociologically, “intuition” is not innate but socially learned, shaped by socialization and repeated exposure to particular worlds. For people who have used Craigslist for years or who grew up on list-based sites and bulletin boards, Craigslist feel natural. For them, its information architecture is “good” because it matches their existing knowledge or know-how. However, for others, such as newcomers to Craigslist, people with different mental models, people who don’t know the site’s grammar, the same information architecture is “bad” because it asks them to climb a steep learning curve. It has failed to include all its users and only privileges those users who are familiar with its interface or in other words familiar with its grammar. Craigslist’s minimalism also serves the purpose of increasing accessibility to users with different abilities. Text-first pages can be easier for screen readers. Consistent layouts aid orientation for users with cognitive disabilities. At the same time, a purely list-based information architecture can be a hindrance for those users for whom headings aren’t descriptive enough. Reliance on list-based information architecture creates a language barrier for multilingual users and anyone unfamiliar with U.S.-specific terms.

So is Craigslist’s information architecture good or bad? The honest answer is both according to the Sociologist in me, depending on who you are and what you already know. It is good at being stable, legible, and fast for repeat, goal-directed tasks performed by people who share its mental model. It is bad at welcoming new users and supporting exploratory tasks. Information architecture doesn’t just arrange content of a digital space; it arranges power: who can find, what can be found and at what cost. Craigslist shows how an information architecture can be both brilliant and brittle: brilliant for those who share its logic, brittle for those who don’t. The task for designers is to preserve the brilliance while softening the brittleness, so intuition isn’t a privilege for the few but a public resource. Its information architecture is “good” because it is stable, simple, and efficient for “learned” use. It is “bad” where that efficiency is at the expense of excluding people who haven’t learned its logic or grammar and the code of efficiently operating it. Therefore information architecture should reflect familiarity for all the users.

Good information architecture leverages existing mental models to build “intuitive” interfaces. However, mental models aren’t universal. Designers inevitably privilege some and sideline others. So whose models get built into a digital space? Usually those of people with power and proximity to the design process: decision-makers, legacy “core” users and the segments that drive revenue. This is amplified by treating frequent, returning users and those with high digital literacy as the norm. In short, the “default user” is a social construction, not a neutral fact. This is where information architecture maps onto social architecture: categories, labels, and pathways mirror existing hierarchies about who counts, who is “typical,” and who is an “edge case.” Sociology helps name and analyze those hierarchies — showing how class, race, gender, disability, language, immigration status, and technical abilities shape whose mental models are centered and whose are ignored. Practically the solution to diagnose this problem means widening whose mental models inform the information architecture which means recruiting beyond convenience samples, co-designing with first-time, low-digital-literacy, multilingual, and assistive-tech users, vetting labels with the communities and treating so-called edge cases as primary use cases. When we design this way, we’re not just arranging content — we’re making the information architecture more accessible and universal.

Mapping the Room: Content Inventory as the First Step to Better Information Architecture

Why Start with Inventory and Not Wireframes

I began our “re-architect the UT Canvas homepage” project for class assuming we would jump straight to the creative and fun part: wireframing. Instead, we started with a very tedious part — project inventory. A project inventory is a complete, factual picture of everything that is in the digital space we are re-architecting. In our project, it took the form of a spreadsheet that lists everything students and instructors actually see and click. Whenever I thought about actually doing the project inventory, my creative brain dialed down a notch. However, I knew that before you re-architect the UT Canvas homepage, you need to learn its landscape. The inventory is how you do that.

Draw the Borders

Early on, I felt overwhelmed by the sheer sprawl of Canvas. If you’re wondering how to start this colossal endeavor, start by drawing the borders. An inventory is only as useful as its scope. Write one sentence that names the boundaries: what pages are in, which are out, and how many clicks beyond the homepage you’ll go. Once the borders are set, decide what counts as a unit of analysis. A link? A card? A banner? Consistency here makes everything easier and cleaner downstream.

Build the Sheet

Use a spreadsheet or table. Give each item its own row and capture what’s on the screen exactly as a user sees it: the verbatim label, where it lives, what it does on click, and where it goes. Add a few more fields that would pay off later: the audience (who it’s for), the owner (who maintains it), the lifecycle (evergreen or time-bound), the status (current, redundant, outdated), and any other metrics. Try to link a screenshot so you have evidence. This step felt tedious and redundant until I discovered a few outdated links and other issues that clearly affected the page’s usability.

Two Perspectives

Now look at the digital space from two different perspectives: a newcomer’s and a power user’s. From the newcomer’s perspective, try to make no assumptions and just pursue common goals and tasks you need to accomplish in that digital space. Then switch hats and browse like a power user who knows Canvas’s “grammar.” When something delights you, note it. When something confuses you or causes friction, note that, too. The goal is to witness how the current structure behaves under different levels of familiarity. This simple step keeps you from designing only for insiders.

Let Patterns Emerge

As your table grows, it turns into a picture. Let the structure of the digital space reveal itself through the inventory. Patterns will surface — like duplicate links pointing to different places or critical actions buried below the fold.

When to Stop

When to stop? Stop when it’s useful. You’ve reached “useful” when you can read your table and name a handful of clear, structural issues without guesswork such as duplication you can eliminate, outdated content you can retire. Later in the project, we would translate these observations into structural moves that would make the page more usable.

Common Pitfalls to Avoid

Don’t let the scope blur. Don’t conflate inventory (what exists) with audit (what to do about it). Separate the steps so the second is grounded in the first. Don’t skip screenshots. And don’t treat the inventory as a one-and-done artifact instead treat it as a living document. Update it when new components appear. A living inventory becomes the backbone of good information architecture.

Impatience → Clarity → Relief

Content inventory is like mapping the room before you begin the creative task of interior designing. You can’t directly jump into designing without first mapping the area. Similarly, you need to first do project inventory before you make any wireframes. It helps you see what’s there, name it precisely, understand who it serves and how often it changes, then let those truths guide your restructuring of the digital space. I moved through three emotions during this task — impatience, clarity, and relief. Impatience at the slow start and the tedium of documenting everything the user sees; clarity as the structure finally surfaced; and relief when the solutions became obvious. The spreadsheet eventually turned chaos into structure.