AI4LAM
COMMUNITY CALL

November 17, 2026 (16:00 UTC)

PART 1: AI4LAM Innovation Award Recipients: Language Identification for Low-Resource Languages in Public Media Archives: A Community-Centered Approach (Yangyang Chen, Rebeca Fraimow, Juliana Principe Salazar)

PART 2: Designing Accountable AI-Assisted Interpretive Pathways for Cultural Heritage (Andrew Ahachinski)

Register here

In the first part of the Community Call we will be delighted to present the third of the four winners of the first-ever AI4LAM Innovation Awards. Established to recognize excellence in the application of AI within the GLAM sector, the awards highlight projects and individuals whose work is helping shape the future of cultural heritage.

In the second part of the Community Call, we will explore a design and governance framework for AI-assisted cultural-heritage interpretation with provenance credits. 

PART 1:
2026 AI4LAM Innovation Award Recipients – Language Identification for Low-Resource Languages in Public Media Archives: A Community-Centered Approach

Authors: Yangyang Chen, Salamasina July Fuimaono, Rebecca Fraimow, Kyeongmin Rim, Owen King – Fantastic Futures 2026 – ConfTool Pro – AdminPapersDetails

Public media archives contain valuable recordings in Indigenous and low-resource languages, yet many of these materials remain difficult to discover because catalog records do not identify the languages spoken. Automatic spoken language identification offers a potential solution, but existing tools often perform poorly on underrepresented languages such as Samoan and Central Alaskan Yup’ik.

In this presentation, we share our experience developing open-source language identification tools for the American Archive of Public Broadcasting (AAPB), in collaboration with language fellows who bring language expertise and cultural knowledge to the project. We discuss how their feedback has shaped our data annotation process, from selecting archival materials and defining language labels to designing a lightweight, browser-based annotation interface.

Rather than focusing primarily on model performance, we highlight the practical and social dimensions of this collaboration, including working across geographic distances, adapting annotation workflows through fellows’ feedback, and navigating community-specific considerations around language data. We also demonstrate our annotation workflow and share lessons that may be useful to other libraries, archives, and museums seeking to improve access to multilingual collections.

PART 2:
From Disclosure to Return: Designing Accountable AI-Assisted Interpretive Pathways for Cultural Heritage

AI tools now write summaries, labels, translations and pathways for heritage collections, in fluent prose that hides where each statement comes from. A notice saying content “may contain errors” tells visitors little, and says nothing about what happens when they disagree. This talk presents “disclosure to return”, a framework of four linked practices for AI-assisted interpretation: say what an output rests on (a specific record, catalogue metadata, a licensed text, a blended dataset, or nothing at all); state what is uncertain and who answers for it; give visitors and communities a route to question or correct it; and show them what changed as a result, crediting them as they choose. The framework is a conceptual proposal, presented at EuroMed2026 in November. It reports no user study, and the talk invites AI4LAM members to test it on their own projects.

Andrew Ahachinsky

is an artist and researcher in Vancouver, Canada, working under the name Bobbinet on participation and credit in the age of AI: who gets to take part in cultural life, and who gets the credit for what is made there. He brings thirty years on stage, as an actor and as a host of community festivals, to the public conversations he now designs. In his format, the Room, the people who made something decide together how each human and AI contribution is credited. His editorial object Pasaporte de Forma was shown at PRINT 2026, Design Week Mexico.