Ground truth

Do statistics provide a sort of ground truth, anchoring our perceptions of the world?

Last year I was invited to participate in a conference on digital innovation in mental health, run by the excellent Dr Becky Inkster. The theme was “ground truth”, that is, the perceptions and experiences of each individual and how they feel about their lives. In the end, I wasn’t able to participate. But the idea of ground truth has sat with me, and I’ve wanted to explore further whether there is any mileage in thinking about statistics as one source of ground truth.

What I mean by this is that statistics provide a lens for us to understand our world, our environment: how good schools are, how expensive houses are, how dangerous society is, how many cases of a pandemic there are in our neighbourhood.

We usually tend to think about these things as products of conscious, active user engagement – in other words, we envisage that someone wants to know about the quality of a school so they obtain the statistical information that can inform them. This type of use is active, visible and incredibly important. Indeed a common theme of OSR reports and reviews is the need for Government bodies to think more about and engage with these types of uses and users.

But I’ve started to wonder whether there is a less visible type of use. Everyone builds a narrative of how the world works based on their personal experiences and information they encounter every day, some of which will come from intermediaries like journalists and social media. This information contributes to building a picture in our minds about ‘how things are’, like assembling a jigsaw puzzle. Hearing a reference to some statistics on the news may lodge something in a person’s mind – not because they’ve actively studied the issue, but because they’ve picked it up in passing.

This type of use of statistics would be very difficult to research. What role might be played by “motivated reasoning” – in other words, how it works if someone hears or reads something that they agree with, as opposed to what they don’t agree with? And how could researchers access these confirmation biases around information that is casually as opposed to purposively picked up?

We have published a literature review on the public good of statistics. We wanted to see what research had taken place on this concept of the public good of statistics – a common phrase that features in the Statistics and Registration Service Act that set up the UK Statistics Authority. My colleague Mary Cowan has summarised the findings in her review, but in short, we found some good empirical work on what people say about their confidence in and use of statistics; some attempts at quantification of value, particularly for economic statistics; and some broader approaches based on a social capital notion of public good.

But all of these approaches tend to rely on a conscious use of statistics – a user who is aware of statistics and actively uses them in choice-making. We found much less on the contextual use of statistics, where a broad awareness of a topic (how is the economy doing) is informed by a range of observations, including statistics picked up in passing through intermediaries, like broadcast news.

It’s possible that there is nothing in this – that the idea of a ground truth role for statistics is misplaced. Or it may be the case that it’s difficult or perhaps impossible to research effectively, or least not possible without extensive participation observation studies.

But we are interested in it. If you know of any research that opens up this area of passive use of statistics, and how it influences perception, let us know!