Using AI in Research Without an Authorized Interpretation Framework Costs Leadership Its Authority
High-stakes research initiatives typically have a large volume of data that lends itself to multiple interpretations. In one project, I noticed that a few focus group participants were circling around a theme that fell outside of our interpretation framework. It wasn’t prevalent enough to draw strong conclusions, but it was enough to raise a judgment call about which data points would inform our next steps—and it had culturally relevant implications for our final outputs. I recommended we expand on that theme in a subsequent quantitative study that revealed new insights we would’ve never had if we didn’t expand on that qualitative theme.
Research organizations across the civic engagement space are navigating a challenging funding climate that has reduced their capacity to carry out their work. Many organizations have laid off staff to stay afloat, and the staff that remains must maintain their output. To solve this capacity problem, some have adopted artificial intelligence (AI) in the research process, especially for the more time-intensive research activities: synthesizing qualitative data. The idea is that an agent can take the first pass at analysis, freeing up a human to take on higher-value tasks in parallel. If an AI-assisted employee can accomplish more, then the capacity problem is solved. The institution believes it has delegated time-intensive and easily automated work. It has actually delegated interpretive authority.
Interpretive authority to define what constitutes a finding in a given cultural context belongs to leadership. An agent lacks cultural competency, cannot infer stakeholder needs, and cannot place the research in a broader context with the progressive movement. If I had delegated analysis and interpretation to AI, the data point that led to a shift in our research direction would’ve been ignored. Without an interpretation framework established before execution, leadership delegates its own authority over the interpretation to AI.
When an organization releases “findings” that are wrong or contextually inappropriate, accountability still lands on leadership. Leadership must publicly defend research insights that were not developed within an authorized interpretation framework. In the civic engagement space, introducing AI into the research process without a defined interpretation framework risks reputational exposure.
Before AI even enters the research process, leadership must define the interpretation framework for analysis and establish governance for the use of AI. Leadership determines what counts as a valid finding. Authority over interpretation belongs to leadership.
Research governance is a leadership function that is formally contracted before an initiative launches.