AI and humor intersect in a groundbreaking study that aims to enhance communication for AAC users by developing AI-powered interfaces facilitating humorous expressions. This research reflects a novel approach to tackle the unique challenges faced by individuals with speech disabilities, particularly when trying to insert humor into conversations.

The study’s primary focus was on Augmentative and Alternative Communication (AAC) users, who often find themselves struggling to deliver timely and expressive comments. AI tools were leveraged to experiment with different design strategies to enable these individuals to infuse humor effectively in social interactions despite their natural communication pace trailing behind that of typical speech.

Several qualitative interviews were conducted with AAC users, providing insights that guided the interface designs. The conversations revealed a significant communication asymmetry, with AAC users communicating at about 12-18 words per minute compared to 125-185 words per minute for non-AAC speakers.

To counteract these challenges, researchers incorporated Large Language Models (LLMs) to predict words and generate contextually relevant humorous suggestions. Four AI-powered interfaces were crafted, each varying in user control and efficiency. This experimentation offered a dual perspective: speeding up conversation rates while ensuring users still delivered personal and relevant humor.

A user study featuring five AAC users tested these interfaces. Participants navigated interactive setups requiring varied levels of input, from selecting keywords to automatic joke suggestions. The study revealed that while some interfaces delivered jokes with speed, others afforded users more control over their expressions.

This research not only underscores the transformative role AI can play in enhancing communicative expressivity for AAC users but also emphasizes the importance of humor in building social connections. The study concludes that achieving a balance between spontaneity and personal expression in humorous interjections is paramount, a finding that holds promise for future interface designs. Looking forward, the aim is to craft an AI-powered tool adaptable to individual preferences, allowing spontaneous and authentic interaction fully driven by the user.