Shiny New Objects
One of the hardest things for anyone who’s in a position to make strategic decisions ever has to deal with is recognizing his or her own decision-making biases. It’s very difficult to step back and ascertain whether or not you are making a decision for the right reasons. One of the most common wrong reasons to choose one path or another is recency bias. You want to be quick to react, and to do so - successfully - ahead of your competition. But you don’t want to react and plot a new course based solely on what’s happened most recently… it may not represent a real trend.
Artificial Intelligence and Machine Learning systems are increasingly able to assist decision-makers by evaluating data points and making a determination as to whether or not they represent a real signal or simply noise. Popular social sites like reddit.com use heuristic algorithms to determine what’s a long-term top topic, and what’s simply hot at the moment. These are excellent boosts for avoiding recency bias and filtering out noise, but they rely on an even simpler fact at their root:
One event isn’t a trend.
I listen regularly to the politics podcast by the folks at FiveThirtyEight, and they too caution listeners and pundits alike against getting caught up in the hype that might surround a single poll. There can exist outliers, and only once you’ve evaluated a poll against other similar polls from around the same time period, can you start to tell just how much it represents a changing trend or a noisy one-off.
I used to tell my team this (admittedly oversimplified) guide: One is a fluke, two is a coincidence, three might be a trend. I’m sure I’m not the first to say something like it, but it’s been a helpful guide. If a data point doesn’t show up more than once, it’s probably just noise. If it shows up twice, it’s interesting insofar as it’s coincidental, but it’s still, likely, just noise. However, if it shows up three times, then it might be worth looking into. It’s a rule of thumb, certainly, but it helps avoid getting overexcited or spending too much time on what could amount to a wild goose chase.
My little aphorism, and even AI and ML, aren’t going to stop every bad decision driven as a result of recency bias or paying attention to noise. But they can help leaders to be more data-driven, and bring these incidents down.