From Interface Accessibility to Universal Design

• CSR
• Design
• Digital equality
• Machine learning
• User experience

Share on Twitter : From Interface Accessibility to Universal Design (New window) Share on Linkedin : From Interface Accessibility to Universal Design (New window)

Since June 2025, European companies offering products and services have been subject to a new accessibility act. Amending the code du commerce (French commercial code), this new European directive means that accessibility is now incorporated into CE marking. This means that accessible design is no longer just a social issue for companies, but a legal one.

As part of the Employee Accessibility research program by Orange’s Mission Insertion Handicap, Orange researchers have been working since 2003 on Universal Design, an approach aimed at structuring the design of products and services, symbolized by the pictogram Figure 1.

Figure 1: a pictogram representing Universal Design.

Current interfaces do not really reflect a principle of equality, rather one of uniformity. The key lies in offering interface variations according to the usage modes.

The Issue of Design in Human-Machine Interaction

Let’s start with an undeniable fact: Human beings are numerous and diverse. This diversity comes from a complex mix of biological, historical, environmental, and cultural factors, which have evolved over time into the variety we see today. While diversity seems to be important on a machine level too (computers, telephones, routers, different OS, different interfaces), it is still limited. This gives rise to a key question regarding human-machine interaction: How can human beings, with their infinite diversity, interact with systems that, though numerous, have limited adaptability?

Humans and machines interact via an interface, which acts as a membrane between the two. This membrane can come in many forms: a screen plus a keyboard, a standalone touch screen, a sensor-lined room, a screen and a remote control, or simply a speaker and a microphone. Whatever form it takes, we need to be able to perceive it (“hmm, what is the system suggesting to me?”), understand what it is offering (“it’s easier in my language, my reasoning is more simple”), and act on it by going through the suggestions (“don’t you have anything else?”) and issuing commands (“do this for me”), as illustrated in Figure 2.

In all cases, the user must be able to perceive, understand, and act on the proposed/selected information.

Figure 2: perceiving, understanding, and acting via an interface. 

For users with specific needs, such as people with disabilities, additional interaction techniques are used, known as assistive technologies. These are interface upper layers that enable access for people with disabilities and work by reinterpreting the initial interface layer (Figure 3). They are generally specialized for a type of disability (visual, hearing, motor impairment, etc.). However, such “reinterpretation,” based on technical criteria, complicates interaction. Typically, screen readers introduce their own navigation logic as well as specific gestures that users will need to learn. This limits the use of assistive technologies to people who can invest a lot into learning how to use them, which can exclude many users such as those with limited technological experience or older people.

Figure 3: perceiving, understanding, and acting via assistive technologies.

Note that the initial layer (the default user interface) is almost always designed within the dominant visuocentric paradigm, where design consists primarily of visually organizing elements in a logic based on pointing. However, to compensate for certain disabilities—particularly motor and visual—assistive technologies must reinterpret the interface in a sequential, non‑pointing logic. And this is where things become more complicated, because some users cannot see, or can see but are unable to point, or have limited vision and therefore point differently, or can point a little but cannot move much, or may struggle to understand that certain objects are designed to be pointed at. Cognitive diversity (neurodiversity) also illustrates the extent of such variations. For example, people with ADHD will be more sensitive to visual clutter and distractions — studies show that they experience far greater perceptual interference when user interfaces are overloaded. Different needs mean different interfaces are required. Cognitive variations can also be seen by comparing an inexperienced user and an expert and their prior experience. In this case, variation can be seen for the same user over time, meaning that the interface needs to offer different solutions over the course of use.

Therefore, the way to design interfaces that are easier to use and suitable for all is by incorporating assistive technologies. In other words, all interfaces should include assistive technologies by default. To achieve this, it is essential that interfaces are multimodal. This means that they provide information in multiple formats (audio, visual, touch, etc.), their content is available at different levels of complexity (some very simple, other content more detailed), and there are various ways of interacting with the interface (voice commands, pointing, via focusing on something). However, even this is not enough. In addition to being multimodal, these interfaces must also be adaptable and adaptive [1][2][3], as illustrated in Figure 4.

Figure 4: being able to choose how you perceive, act, or are assisted by the machine.

This isn’t just about changing a background picture, it’s about how to bring all of this together into a single adaptable interface that accommodates all these perceptual, cognitive, and action-related constraints. A line of research conceptualized this in the 2000s with the term “Universal Design” [4][5].

Toward Universal Design

According to Article 2 of the UN Convention on the Rights of Persons with Disabilities, “Universal design” means the design of products, environments, software and services to be usable by all people, to the greatest extent possible, without the need for adaptation or specialized design.” In practical terms, implementation has however been limited as this cannot be achieved through a single interface, only through multiple. But let’s get to the heart of the matter. While a single, average interface that is a compromise for everyone would of course be a commendable effort, it would not be enough. For example, some users, such as those who are visually impaired, would want to have audio, whereas others definitely would not, such as users who are hypersensitive to sound. Similarly, some users need a blue background with yellow text, while for other users this would not be suitable at all as they are the exact colors that they cannot perceive. What’s more, an interface must not be something that is chosen by the majority and imposed on the minority, nor—as often is the case in standardization—should it mean the quietest voice needs to compromise. It must allow for customization, being tailored to the needs of each individual. The pinnacle of personalization.

1) A Digital Republic

It’s rather interesting to see that, even before ergonomics comes into play, the matter is primarily political, because digital interfaces have become an indispensable part of social life. The interface that we are talking about here must reflect both uniqueness and diversity at the same time. In France, inclusion of differences is underpinned by the national principles of equality, fraternity, and liberty. It is important to bear in mind, however, that equality does not mean uniformity. Thankfully, we are free to be different. Liberty is ultimately about being able to express that difference, and above all equality means having the right to that difference.  We should remember that the 2005 French law for the equal rights and opportunities, the participation and the citizenship of persons with disabilities, has already established the principle of generalized accessibility, including all disabilities, whether physical, perceptual, or cognitive. It is therefore a person’s right to have the needs presented by their disability addressed, regardless of the origin and nature of the impairment, their age, or their way of life. However, up until now, accessibility of interfaces has been based on a single, common interface for everyone. It has been up to people with disabilities to use their assistive technologies, which are assumed to be compatible with this common interface — something that is rarely in fact the case. A 2025 study by the Contentsquare Foundation revealed that 94% of e-commerce sites do not have accessible payment pathways. More generally, it is estimated that less than 10% of websites in France comply with the accessible design rules as defined by the WCAG (Web Content Accessibility Guidelines) 2.2. Such accessibility and ergonomic shortcomings contribute to digital exclusion. According to the latest report of CREDOC (Centre de recherche pour l’étude et l’observation des conditions de vie — French research center for the study and observation of living conditions) (Baromètre du numérique, édition 2025 — 2025 digital barometer), a third of the French population says they encounter obstacles with using technology in their daily lives. We are not really operating on a principle of equality, rather one of uniformity. Of course, a common basis must exist. But it must be flexible enough to be adapted. What we need is a customizable interface that we can configure according to our own differences — an interface that upholds the right to that difference.

2) Usage Mode Approach

Taking diversity into account in a user interface means allowing everyone to have their own interface adaptation to interact with on their own, in the way that suits them, without being excluded from interaction. As mentioned, many different needs exist, meaning many potential adjustments are necessary.

The concept of usage modes [6] aims to address this problem. Usage modes are preconfigured settings tailored to very specific interaction methods, designed to ensure equality. By choosing a usage mode, the user selects an interaction style corresponding to their needs, which automatically determines an appropriate configuration. They therefore do not need to carry out complex adjustments to their interface when using it for the first time. Each usage mode corresponds to a specific optimization related to a type of disability or difficulty (for example, a person who is new to touch interfaces, is experiencing reading difficulties, or has limited motor skills in their fingers). In this way, optimization aims to reinforce the methods that the user utilizes in their daily lives. For example, for a visually impaired user, this would involve enhancing the interface’s graphics (contrast and element size) as well as making the interface more reliable (for instance, by extending the press time to confirm an action). Orange researchers offer the following seven usage modes that address the main categories of accessibility while remaining easy to manage, namely:

Easy+ : for simplified use, suitable for people with little technological experience or with cognitive difficulties.

Visual+ : with visual enhancements, for people with visual impairments.

Vocal+ : for people without sight or those who prefer sequential or voice navigation.

Motor+ : with sequential navigation options using the keyboard and other peripherals, for people with motor difficulties.

Pointing+ : to facilitate pointing, useful for people with fine motor impairments.

Reading+  : to improve readability and help with reading.

Audio+ : with subtitles and other aids, for people with hearing impairments or specific hearing needs.

Several prototypes have been used to test the majority of these usage modes, including the Tactile Facile mobile app (Figure 5) and the Comfort+ browser extension (Figure 6). Presentation, content type, and available interaction options therefore differ from one mode to the next.  Other usage modes are being developed in these two prototypes, including a Serene+ mode for individuals with sensory hypersensitivity and a sign language mode for deaf individuals who communicate in sign language.

Figure 5: Tactile Facile, a prototype implementing Universal Design; a) the usage modes setup screen; b), c), and d): the home page for the Easy+, Vision+, Graphic usage modes.

Figure 6: Comfort+, an open-source browser extension to aid reading and navigation: confort-plus.orange.com/index_en.html

In short, as needs vary widely and there are far too many settings, usage modes have been created to offer users ready-made settings bundles. But this is still not enough, as it is a little too static and fails to take into account users’ usage habits.

3) Interface Adaptability and Customization

Users have established usage habits. They follow specific strategies, whether these are shaped by visual or auditory constraints, or limitations related to motor skills or comprehension: starting again from the beginning, listing hyperlinks, reading only certain elements, relying on the margin or capital letters, etc. Each of these strategies requires different tools and features. This means that settings often need to be fine-tuned; however, simplicity still remains key. How about if the machine could do some of the work? It would first need to filter for the settings for a particular individual and, even better, suggest settings adjustments to that user based on their activity. This is where algorithmic and software engineering comes into play, as illustrated in Figure 7.

Figure 7: an abstract hierarchical model, generic components, usage modes, continuous machine learning, and adaptation rules. 

For interfaces to automatically adapt, however, we need AI that can learn continuously in a changing environment. AI that is not based on the data from a group of individuals, rather based on a much more limited data set — that of the current user. AI that can analyze an individual’s usage habits, with all their particularities. Orange’s ABIT algorithm [7] can predict all of a user’s actions and action sequences. To do this, a small snippet of code is deposited on the interface’s key elements (buttons, menus). This piece of code feeds the algorithm, which can then generate probabilities about the future, based on past usage and actions. With very little data, this algorithm can create a task tree (Figure 8). Each level of the tree represents an interaction level, or screen. Between the levels, the calculations of the conditional probabilities of actions are shown, based on the actions previously performed. Very probable pathways (in color in this diagram) appear. The interface can then suggest things to make the user’s life easier, simply guiding them by saying (visually and/or audibly): “Hey, last time you went this way. Do you want to follow the same pathway again?”. This can be helpful for people who tend to always do the same thing and have a bit of trouble finding their way due to the interface’s complexity.

Figure 8: a task model reflecting the logical structure of the application and predicting possible actions at a given time.

We can take it a step further, however, and explicitly suggest creating macro commands: “Hey, you often do this series of ten actions, what do you think about me creating a shortcut for you that will do them all at once?”. And where probability is as certain as it can be, the system could even suggest doing these ten actions automatically: “Can I suggest making things easier for you by automating this task, which appears to be exactly the same every time?”.

Conclusion

Universal Design is centered around three complementary principles: integrating multimodality by default, offering users a choice of usage modes, and providing customization and adaptability. This approach can be described as utopian, because a utopia is a distant goal — a dream of going beyond current limits, often considered unattainable. There is also the concept of concrete utopia, developed by Ernst Bloch [8], a transcendence that materializes in a tangible form, inscribed in the materiality of the world. Universal Design can therefore be considered a concrete utopia as it is rooted in values that illuminate and structure the digital world of today and tomorrow, and yet Orange’s research has proven it feasible through several prototypes. This, however, requires a radical change in design processes. The first radical change is to orientate design to the users’ needs, something that ergonomics and design have been attempting to do for many years. The second is to structure this ergonomic design around interface variations by usage mode. The third and last, but certainly not the least, change is for design to be highly customizable, where it can be adjusted at first usage and as the individual continues to use the interface. It certainly seems an easier process than getting to Mars, so shall we give it a go?