All AI-assisted work at Ashoka is governed by Ashoka's Acceptable AI Use policy. The Policy focuses on the appropriate use of AI Tools and their associated outputs for the purpose of protecting Ashoka’s proprietary data and the privacy of our community of changemakers; ensuring compliance with applicable laws and regulations; and defending Ashoka’s reputation as a trusted source of insights and analysis of system change ideas.
In alignment with our core vision of an Everyone a Changemaker (“EACH”) world, we are committed to the responsible use of artificial intelligence (“AI”) as we explore solutions that harness the potential of AI for the good of all while expanding Ashoka's capacity to amplify EACH. AI Tools can be employed to serve many supportive purposes, such as digesting information about a topic, recommending workflow improvements, or inspiring creativity. At the same time, use of such tools also involves risk, both in connection with the information provided to the tool (all of which could become public) as well as in how the generated content is then used (e.g., informing hiring decisions). Ashoka's Acceptable AI Use policy provides guidance on how to avoid and/or mitigate such risks. Given the ever-evolving nature of AI, Ashoka will continually assess the Policy and update it according to emerging standards, best practices, and use cases.
Transparency - Tech for Human Agency apps leveraging AI
In addition to these safeguards, within our Tech for Human Agency work, we are committed to transparency in two ways.
First, we publish a sIfA graph (Statement of Intellectual Fellowship and Accountability) for each Tech for Human Agency tool that leverages AI. sIfA asks questions about how humans have collaborated with each other first, followed by questions on how each contributor has interacted with AI. The tool creates both a table as well as the sIfA figure: a radar diagram in the shape of a brain with fluid lines to express that human/AI interaction is not clear cut, even when declared. The sIfA figure can be used as a visual to foster a culture of transparent reflection and learning when it comes to interactions between humans and machines.The sIfA tool can be used and downloaded at www.busara.global and on GitHub at https://github.com/mareikeschomerus-ctrl/sIfA.
Second, we strive to develop app specific metrics to score our Human-AI collaboration and establish a high-enough threshold to ensure humans are actively exercising judgement rather than rubber-stamping AI outputs.
| Tool | sIfA graph | Human-AI collaboration narrative & score |
| Tech for Human Agency Pulse | Tech for Human Agency Pulse - sIfA | TxHAPulse is produced through a structured editorial workflow where AI generates candidate stories from Fellow social media and public sources, and a human editor reviews, filters, edits, and publishes the final digest. We measure this collaboration using a 60-point score across three dimensions: selection filtering (how much the editor curates the AI-generated candidate pool), tier decisions (prioritisation of stories), and text editing (extent of rewriting). The AI contributes a fixed 40-point baseline for sourcing, drafting, and structuring; the human editorial score reflects the judgment applied on top. Current score: 40.2 / 60 — Active editorial oversight. Read the full methodology |
We may not get everything right on the first try. Do reach out to use at techandhumanity@ashoka.org for questions or feedback.