Searching for Needles in Haystacks: How our brain sabotages our searches

In a recent study funded by the U.S. Department of Homeland Security (DHS) and reported in LiveScience, researchers found that subjects’ expectations of finding something had a direct effect on their success rates for finding the items in question.

Found it yet?

Found it yet?

In the study, subjects looked at X-ray scans of checked baggage and tried to identify the presence of guns and knives.  In the first trial, a gun or knife was present in 50% of the bags, and subjects only missed the weapons 7% of the time.  In the second trial, the guns and knives were in only 2% of the bags, and the subjects missed the weapons 30% of the time.  In short, when something is harder to find, our accuracy in identifying it drops significantly.

This is a trick our brain is playing on us as it becomes bored when we do not find what we are looking for and stops paying attention, meaning we then miss things when they do appear.

While the implications for airline security are obvious and somewhat chilling, the implications for the enterprise are also worth examining.  Knowledge workers spend ca. 15% of their day searching for content.  Applying the lessons learned in the DHS study, we can assume that if a search query returns fewer correct results in relation to incorrect results, the knowledge worker’s accuracy in picking out the relevant items will decline.

Conversely, just as in the DHS study, if the correct to incorrect ratio is better, meaning there is a higher number of correct results, then the knowledge worker is much more likely to find more of them.

For knowledge-based organizations and providers of software to these groups, the lessons from this study are clear: search tools must be improved to provide better ratios of relevant, useful results.  Today’s search tools focus on returning large sets of results and the answers to a search query may very well lie somewhere within these.  However, the low signal-to-noise ratio virtually ensures low accuracy even if one were to comb through every last result.

Search results need to be highly contextual and limited in volume to ensure accuracy and provide a favorable ratio of correct to incorrect results.  This keeps the knowledge worker engaged and not feeling that he is looking for a needle in a haystack; this, in turn, increases the probability of identifying the needed content.

Cody Burke is a senior analyst at Basex.

Comments are closed.