AI Policy - PROPOSAL FOR CONSIDERATION
Our value to the world rests on one thing above all others: that people can trust our historical record. Every use of AI at the Society either serves that trust or does not happen.
Artificial intelligence at the Society is a research assistant working under human direction. It does not speak for us, it does not decide what is true, and it does not stand in for the people who make up our community.
1. Ethics Control
Author credit. Our mission is to tell untold stories and to credit people whose work was overlooked; for that reason, clear author credit is a priority.
Human authority. Every registered Historian on our team may reject or revise any content in our library (AI-assisted or otherwise).
When in doubt, ask. Anyone unsure whether a use is appropriate should pause and ask before proceeding. Questions can go to committee members, a committee leader, or to the Board; raising a question is always welcome.
AI and our members. We do not use AI for member communication, and we do not place a chatbot between ITHS and its members. As a society, we exist to foster human-to-human connections among members and the public at large.
2. Quality Control
An accurate and clearly sourced record. Trust depends on accuracy and on clear provenance. The Society does not publish text drawn from unknown origins.
Training data and provenance. We do not rely on generative output of unknown provenance to establish a primary historical fact.
Encouraged uses. Using AI tools for settled, non-subjective work, such as checking grammar, confirming a matter of fact, or classifying an object, is encouraged.
Copyright and permissions. We respect the copyright and licensing of images, text, and archival material, and we remove items of dubious permissions in favor of material with clear usage rights.
Fair use of publicly visible works. As historians we sometimes describe and summarize works which are available to the general public. We hold that using a software tool to help prepare such summaries is no less legitimate than an human performing the same task. In each case we credit the source by name and by link, Our aim is to point to the work and give credit, not to reproduce it.
3. Standards of Practice
Audit trail. The Society keeps an internal, auditable record of which people directed which AI-assisted processes. AI-assisted tools note what they changed and why, so that any historian can trace how a piece of data came to be and, if an error is found, check for similar errors elsewhere.
AI as a research assistant. We use AI for the time-consuming, repetitive, high-volume work that a small volunteer team could not otherwise complete: finding and repairing dead links, classifying and cross-referencing objects, finding sources which support claims made by others, and creating well-cited abstracts which can be relied upon as references back to original works.
Pre-screening submissions. We use AI to pre-screen submissions for malware, spam, SEO or SEM content, self-promotional material, commercial links, and similar problems. Humans have the final word; AI serves as an advisory and screening layer that reduces the burden on our team.
Consistency across volunteers. Because our historians work independently, we keep shared standard operating procedures so that everyone works to the same standard. This is to ensure that the tool working for one historian meets the same standard as the tool working for another.