April 18, 2013

Why First Impressions Count

Test Management
In my last post on thinking and bias, I talked about how Daniel Kahneman’s System 1 and System 2 models of thinking led us to approach situations with biases that could influence how we made decisions about those situations. Now I’d like to delve more deeply into the specifics of some of those biases. One such bias, the halo effect, has been researched extensively and written about by Philip Rosenzweig, who summarized them in his book The Halo Effect. In this book, Rosenzweig looks at the delusions that we commonly experience in our working lives, and how they affect the conclusions we arrive at as we attempt to explain success or failure. The halo effect is all about us wanting our first impressions to be proven correct in our impressions. But where do those first impressions come from? Often they are planted in our minds by earlier conversations, reputation, or other anecdotal method. Whether positive or negative, they influence how and what we test. In addition, there is another aspect to the halo effect. In the absence of any information, we also tend to be predisposed to view a person or situation favorably. We tend to view the situation as positive, and respond to it as supporting the conclusion that all is well. But if the situation or the product turns negative, we go back and reevaluate our evidence with that in mind, biasing our view of the evidence based on the conclusion we already know. A strong, decisive leader becomes remembered as rigid and unyielding as the project sinks into difficulty. How do we apply the halo effect to software testing? Often we are deceived by what we hear about a particular project before we start testing. The team is the best in the company, or an insider has high expectations for the product, so we are likely to be less than aggressive in our testing. On the other hand, if we doubt the quality of the team, or the technical details of the project, we may become more aggressive in our testing. Can individuals and teams overcome the halo effect? At conferences, I’ve been preaching that automation has benefits that go far beyond the usual ROI calculations that many depend on. Automation reduces bias from the testing process. It doesn’t completely eliminate bias because it depends on how we define our test strategy and test cases, but automation itself stresses repeatability. We do the same thing in the same way, enabling us to eliminate human variability from the process itself.