In June, I had the opportunity to attend UX London 2026, one of the leading conferences for the international User Experience community.
Events like this always remind me why I enjoy attending conferences so much. It’s not just about learning new methods or collecting ideas to bring back to work, it’s about challenging the way we think about the people we design for.
As a User Researcher, I spend much of my time trying to understand behaviours, needs, and perspectives that differ from my own. That’s why the recurring theme throughout the day, the importance of seeing the world through multiple perspectives, resonated so strongly with me.
Among all the talks, Michael Kibedi’s was probably the one that left me with the most questions.
Starting with a simple but powerful provocation “Whose English gets to be the default?” Kibedi invited us to reflect on how accent bias has become embedded in the voice technologies we use every day.
How many times have we heard someone say that a voice assistant couldn’t understand them? Or perhaps we’ve caught ourselves changing our pronunciation or repeating a sentence just so a machine could recognise what we meant?
Although the talk focused on English, I couldn’t help thinking about how relevant this is in Italy as well. After all, the Italian spoken in Milano doesn’t sound like the Italian spoken in Napoli, Palermo, or Bari. Our accents reflect where we come from, our histories, and our identities. Yet we often treat certain varieties as more “neutral” or more standard than others. If these hierarchies exist in society, it’s worth asking whether they also find their way into the technologies we design and use.
What I found particularly compelling is that Kibedi doesn’t frame these issues as purely technical problems. Instead, he encourages us to see them as the outcome of design and cultural decisions: whose voices are included in the datasets? Which accents are represented? Which ways of speaking are treated as the norm?
As a researcher, this strongly reminded me of our own daily work. Every time we recruit participants, define personas, or segment users, we’re making decisions about who is represented and who might unintentionally be left out. Voice technologies simply make these choices more visible, but the underlying principle applies to every digital product we build.
When Bias Becomes an Algorithm
Another part of the talk that really stood out was the growing interest in systems that claim to infer emotions, personality traits, or other individual characteristics from a person’s voice.
Kibedi encouraged us to approach these claims with caution.
Can we really infer meaningful aspects of someone’s identity from the way they speak? Or are we simply translating cultural stereotypes into computational models?
Over the past few years, enthusiasm around AI and automation has grown tremendously. Talks like this remind us that innovation cannot be separated from critical reflection about its broader consequences.
The Value of Polyvocal Futures
The talk concluded with an idea that I found particularly inspiring.
Kibedi spoke about polyvocal futures: futures where the diversity of accents, dialects, and ways of speaking is not treated as a problem to be corrected, but as a richness to be embraced.
It made me realise that this idea extends far beyond voice interfaces. In a technology landscape increasingly shaped by artificial intelligence, we should ask ourselves more often: who is represented in the systems we build, and who risks being left behind?
Every design, technological, or organisational decision carries a particular view of the world. Recognising that is the first step toward building technologies that are genuinely inclusive.
Credit to: Graziella Dramisino