Only humans can swipe

Success for an experience designer is creating great products and services that help people do their thing in an easy, intuitive way.

In the digital world, we had just about settled on patterns and principles for successful website and app design, when chatbots appeared. We suddenly had a new interface – language – and everything we knew about interaction was changing. We had to work out how to create a system that can navigate the complexities of conversation to understand and handle a request.

Pütz and Esposito’s (2024) research shows how this challenge is not going away with the next generation of conversational agents: it is just changing form. The one constant, though, is the need for a human to get the conversation back on track when it fails.

Conversation is complex. Requests are messy, unpredictable, and indirect. It might be easy to instruct a voice assistant to understand that the question ‘do I need an umbrella?’ is soliciting an answer about the weather. But what about scenarios like these: Could a customer service agent in a bank understand that when a customer says, ‘I couldn’t get through the train barrier’ they mean ‘my card didn’t work?’ Or would a delivery service adviser know that ‘I’m going on holiday next week’ means ‘will my parcel arrive by Friday?’ And would emergency service operators know that ‘I’d like to order a pizza’ could mean ‘I need urgent help’ (Stokoe, 2018)?

These nuances are easy for people to understand but difficult for an automated system to get right. LLMs have taken conversation agents to the next level, but there is vast scope for them to go wrong. They still need a human to fix them, whether that human is the designer or the user.

Humans are adaptive creatures. Scrolling and swiping were once novel gestures to us, and now they are second nature. People using conversational systems will forge their own path to successful interactions in a similar way; even today, I hear my children relentlessly reprompting Alexa, adjusting how they ask questions until they get the output they want. We might have a design principle that ‘we should never make people change the way they naturally interact’, but people will not wait around for us to solve for that. They will find their own ways to ‘swipe’ in this new interface; they will develop their own new ‘communicative competence’.

Usability expert Nielsen (2023) has said that AI is creating a paradigm shift in the way people will interact with products and services. We are moving from instruction-based to request-based interactions, and the interface of the future will be language. As designers, our goal to make better products has not changed. But as the user interface evolves, we need to adjust our methods and draw on new expertise. Creating successful language-based interfaces means getting closer to the components of conversation; it means leaning on conversation analysts to guide us to success.