How can you make IA more inclusive for diverse needs?

Make AI more inclusive by ensuring diverse and representative data, adhering to ethical principles, involving users with diverse needs, implementing accessibility features, considering cultural differences, and regularly auditing for bias.

Understanding the context of your digital product will help you to understand what you need to know about your users. With any IA your main goal is to funnel users to the star of the show, the content. So the labelling and structure needs to be reflective of how your users need to be funnelled.

A inclusive tone is a key element for having an inclusive AI. This is difference between the type or variant of a language - like American English, British English or Scottish English and the having a defined tonality for conversations. Inclusivity is more about empathy and about evolving a emotional template for interactions and defining a tonality for conversations. It is almost like designing conversation not just with words and weaving it with emotions. It is rooted to the mission and culture of one's organization and shaped by the values of your brand. All these define the choice of words, adjectives, sentence structure and emotional pitch.

This advice is a little paradoxical. While jargon should be avoided in most cases, in some instances it works well with the audience type and the brand. If your digital product is being delivered to a global audience what you *should* be doing is understanding if there's rationale for the site to be provided in other languages serving a specific market. If you're going to provide a global site in English, you need to look at the originator market. If the org is based in the UK, British English, if it's in Australia then Australian English. Otherwise you should be using localised English versions.

You can design with your users by using card sorting to either create or test your labels (open card sorting means your users both sort AND label the content, whereas closed card sorting you're asking your users to sort the content to the labels you've defined - or testing the ones they've created for a double loop test). You can use tree testing to test the structure of information, i.e. do users expect to find information where you've filled it. Please be aware that card sorting is not the best technique to use with people who have visual impairments or some who are neurodiverse as it requires a lot of information to be held at one time.

The biggest tip is don't use the same non-descriptive links across a page. For example, if you've got a landing page with image blocks and titles laid out across the page, don't have 6 links that say "click to view more" or "read more", it's not helpful for people using screen readers.

Labelling navigation structures is one of the most misunderstood elements of design. If you engage your stakeholders in the process, be prepared for them to demand changes without understanding what chaos that will cause. This is something nobody else is going to tell you. Your labelling, and by extension the navigation, needs to be all but invisible to those who interact with it. If your users can't figure out what to select or where to go you've got a problem. They will default to search. If you're lucky that will be site search, if unlucky, it'll be Google, and you risk losing them altogether. Be mindful of what insight you need from stakeholders as they can cause more harm than good.