Explainability in Speech and Audio AI at the Edge

Explainability in Speech and Audio AI at the Edge

Event Description:

With the growing use of speech and audio AI on local devices such as smartphones, wearables, and in-vehicle systems, explainability has become increasingly crucial for trust, accountability, and safety.
However, what makes an explanation meaningful depends heavily on the environment in which the model operates and the people who interact with it.
This workshop brings together researchers to discuss how explainability can be designed for speech and audio AI at the edge—where computation happens close to the user and transparency must coexist with efficiency and privacy.

Info: Welcome by Xiaoliang Wu and overview of aims and format.

Info: Participants collaborate to list real examples of speech and audio AI operating on edge devices.
Each group discusses the function, user context, and potential need for explanation.

Info: Groups select representative applications and analyse them from various stakeholder viewpoints—end user, developer, deployer.
They examine what concerns arise, what explanations are required, and what barriers limit explainability.Session 2.A for first two.

Info: Groups select representative applications and analyse them from various stakeholder viewpoints—end user, developer, deployer.
They examine what concerns arise, what explanations are required, and what barriers limit explainability. Session 2.A for last two.

Info: Participants form new teams for an interactive challenge.
Each team reviews the barriers identified earlier and proposes concise, practical ideas for overcoming them through improved design, interaction, or communication.
Teams present three-minute pitches; participants vote for the most promising ideas.

Know what’s Happening Next — before everyone else does.

Get App

Host or Publisher Xiaoliang Wu