Accessible conversational user interfaces: considerations for design

Conversational user interfaces (CUIs), such as chatbots and voice assistants, are increasingly common in areas of day-to-day life, and can be expected to become ever more pervasive in the future. These interfaces are being designed for ever more complex interactions, and they appear to have potential to be beneficial to people with disabilities to interact through the web and with technologies embedded in the environment. However, to fulfil this promise they need to be designed to be accessible. This paper reviews a range of current guidance, reports, research and literature on accessible design for different disability groups, including users with mental health issues, autism, health conditions, cognitive disabilities, dyslexia or learning difficulties, and sensory, mobility or dexterity impairments. We collate the elements from this body of guidance that appear relevant to the design of accessible CUIs, and instances where guidance presents issues which are less conclusive, and require further exploration. Using this, we develop a set of questions which could be useful in the further research and development of accessible CUIs. We conclude by considering why CUIs could present opportunities for furthering accessibility, by introducing an example of this potential - a project to design an assistant to support students to disclose their disabilities and organise support, without the need to fill in forms.