Minding the gap between data and understanding

While data is raw and unorganised, information is structured and organised, making the latter a better tool for decision-making.

Search engines, such as Google, have been revolutionary in allowing us access to salient information at the moment we need it most. By contrast, social media algorithms, by optimising for clicks and shares, can often end up pushing us data points that play on our emotions, rather than seeking to inform us.

This important distinction may well play an important role in why consumer social media apps could have a very limited role in workplace communication, where optimising for commercial outcomes requires very different thinking.

In Thinking, fast and slow, Daniel Kahneman distinguishes between our system 1 brain and our system 2 brain. Our system 1 brain is fast, instinctive and emotional, while our system 2 brain is slower, more deliberative and more logical.

As humans, when we first approach a situation, our system 1 brain may well be in control until our more rational system 2 brain takes over. While information that plays on our system 1 brain might be more entertaining for us, our system 2 brain may well help us produce better outcomes in unclear situations.

Recently, I found myself working with a team that was building a software training platform to help improve management decisions. The team found it hard to present information in a way that would incentivise managers to learn more about their teams and drive improvements to their performance.

The immediate advice I gave for improving user experience was based on heuristics developed by the Behavioural Insights Team, a company formed about a decade ago from within government to help nudge citizens to make smarter decisions about health, wealth and happiness. One useful mental model is the East framework, where you focus on making decisions as easy, attractive, social and timely as possible.

More advanced frameworks like Mindspace introduce other factors that can be used to nudge behaviour, such as leveraging the fact that people like to act in ways that make them feel better about themselves.

These small interventions can have big effects. For example, the Behavioural Insights Team found that by simply using text message reminders in adult education programmes, there was an 8% point increase in the likelihood of a control group passing exams over an academic year.

When making more complex improvements to user experience at a large scale, especially where prior evidence is more limited, it is important to measure the impact to make sure these interventions are not doing more harm than good. For example, scientifically robust randomised control trials, in which people are randomly allocated into control and trial groups, can help to provide conclusive answers quickly in a large user base.

This can be tough, though, when designing a training programme for a small audience that is trying to move a north star metric that has a slow feedback loop.

This challenge shows just how hard it is to automatically editorialise content or create education curriculums. At the same time, with more data available than ever before, it is essential to be able to distil down the information that matters about a subject. Therefore, editing information to make it as salient and simple as possible will be an ever more important endeavour in our knowledge society.

Junade Ali is an experienced technologist with an interest in software engineering management, computer security research and distributed systems