Tapping into research with Consensus

Have you ever started a project and wished you could tap into the whole world’s research in seconds? Understandably this is something that has come up for us regularly, as we draft out our Enablers of change episodes. We thought it’d be good to talk through what AI we use and how we use it to highlight relevant research and speed up our production process. In this episode, we’ll share what we’ve learned from using Consensus and how it helps bring research into our work.

Consensus is a tool that searches peer-reviewed research and synthesises answers to questions based on evidence. Instead of skimming articles one by one, we can ask a question and the model pulls together insights from multiple studies. It’s designed to surface consensus where it exists, not just highlight the first result from a search engine. Hence the name!

For enablers of change, this matters because research and extension are linked. If we’re recommending an innovation or summarising how a practice works across different contexts, we want to base that on solid evidence. Consensus doesn’t replace full reading of original papers, but it helps us quickly see what the research says and decide where to look deeper.

The first thing we discovered is that Consensus works best when we ask specific questions. If we start with a broad question such as ‘What does research say about action learning?’, Consensus provides a good overview of how action learning is generally found to be an effective way of developing problem solving, professional skills and leadership. However, if we ask: ‘What does peer-reviewed research say about the impact of action learning on on-farm change?’ then we get more relevant research being highlighted, including where there’s been specific practice changes when using action learning, for example, Morgan et al. (2021) pops up on antibiotic use and action learning. This prompt also provides an overview of which part of on-farm change is impacted by particular features of action learning.

By narrowing our focus, we get an answer that summarises relevant studies rather than a mix of loosely related findings. One of the biggest benefits of using Consensus is the speed with which we can check the research! When we want to include a statement in a blog post, especially when we’re exploring a topic that we might not know a lot about, we can run it through Consensus and see what the research says and then read specific papers to see what’s useful.

Our approach has been to decide which area of research we’re trying to understand, and then draft a sentence we can ask Consensus. Our questions tend to follow the pattern of asking what the research says about a particular topic in the context of agriculture. Or we ask whether something is useful for supporting change on-farm. We then review the summaries and check the links to studies the tool provides. We’ll go back to our draft episode and update it to add research to support a particular point or rewrite to reflect what the evidence actually shows. And of course we can then add references to our resource list at the end of the blog post so we can all read more details if we’d like to. 

Consensus can also create a visual meter by analysing peer-reviewed papers relevant to our question, using natural language processing to determine whether each study answers the question and how its findings align (whether that be Yes, No, Mixed or Possibly), then aggregating those classifications into a summary. The “Consensus meter” displays the balance of evidence as a simple bar chart with the number of included studies, along with the references used so we can read more details. For example, in the figure below, 18 of 19 studies (95%) are classified as supporting the conclusion that climate change is overwhelmingly caused by human activities, with one mixed and none opposed, providing an immediate visual snapshot of the evidence base.Based on our experience, we think Consensus is a really useful tool for enablers of change. As always with AI, there are some cautions. Don’t assume consensus means there’s no nuance or qualifiers to the research. Avoid vague queries as they produce a long list of relevant but not necessarily useful papers. And don’t skip reading the actual papers! We always need to check that AI isn’t making up its claims.

Given all this, we like that Consensus helps us easily explore research and ensures we’re providing evidence for what we talk about here. Tools like this don’t replace our judgement, but they expand our capacity to access and integrate research. We can also save a bit of time and improve the quality of our content. 

We’ve shared our thoughts, now it’s your turn! Add a comment below with your experiences, tips and ideas on using AI tools like Consensus to help understand what research might inform extension practice. We’d love this to be a conversation as we value your insights. 

Thanks for joining us for this episode! Don’t forget to subscribe to our newsletter to catch future episodes, and share this post with friends who might enjoy the discussion. 

Resources

Morgans, L. C., Bolt, S., Bruno-McClung, E., Van Dijk, L., Escobar, M. P., Buller, H. J., Main, D. C. J. & Reyher, K. K. (2021). A participatory, farmer-led approach to changing practices around antimicrobial use on UK farms. Journal of Dairy Science, 104(2), 2212-2230.