20/04/2026
What if AI could tell you what every candidate in the Holyrood 2026 election actually stands for before you vote? Wouldn't it be ideal if you could just ask any question about any issue and see how a specific candidate is likely to respond? Or perhaps compare multiple candidates side by side on the same topic and quickly understand where they differ?!
Today I have launched LikelyStance.com to solve a simple but very real problem. Voters are expected to make important decisions using fragmented information scattered across dozens of sources. Manifestos, interviews, debates, social posts and parliamentary records. It is all there, but it is not usable. And so, this new platform is designed to tackle that problem, head on.
LikelyStance continuously ingests public data from a number of sources....candidate statements, party manifestos, local leaflets, media interviews, debates, campaign materials, and parliamentary records. That data is then structured to ensure statements are attributed correctly, and classified across a defined set of political topics. Underneath this is a stance detection funnel that evaluates whether a candidate is expressing support, opposition, concern, or neutrality on any given issue, weighting this based on consistency, strength, and recency of their statements.
The output generated isn't an opinion, it is an evidence based interpretation of what candidates have actually said and what they consistently stand for. If a particular candidate has not spoken on a topic, that absence is reflected in the output. And where necessary, the model uses party positions as a clearly signposted proxy. Crucially, this is not a system that is able to hallucinate or invent positions.
I am very aware that there is an underlying tension here. If I had waited until every candidate had spoken in depth on every issue, this thing would never have launched. So what has gone live today is very deliberately an early version. It is already very useful, however it will get meaningfully better hour by hour and day by day, in the run-up to the election. As candidates say more, as a broader data pool becomes available, and as the model continues to learn from real usage and feedback, the quality and depth of the insights will get better.
A huge thanks to a number of people within my Linkedin network who tested the early versions. You know who you are and it made a genuine difference. If you are interested in how AI can improve political transparency and voter understanding, I would really value your feedback.
RH
