As originally conceived, empiricism was a way to test your theories and hypotheses against observations of the natural world. This approach to understanding reality was a revolution in the sciences, bringing a flood of scientific knowledge and technological advancements.
It’s not surprising then that in modern applied sciences, such as human-computer interaction and software engineering, empiricism plays a similar role in helping researchers understand the phenomena they work with. One major difference between “soft” sciences such as HCI and software engineering and “hard” sciences such as biology, chemistry, and physics, is that the phenomena they study are quite different in their permanence. Physicists study the inner workings of the laws of the universe; biologists study the form and function of processes in life. These are phenomena that change very slowly over time, which gives “hard” scientists time to hypothesize, theorize, reject, and synthesize.
In contrast, “soft” sciences study phenomena that are extremely quick to change, relative to the expansion of the universe and evolution on Earth. I might suggest some relationships between programming environment design and group work, only to have yet undiscovered and fundamental dimensions of these programming environments change under my nose. I might propose a theory about the causal relationships between programming paradigms and productivity, but in 20 years, someone could reinvent the programming language, possibly making my theories irrelevant. Worse yet, by relying on empirical data, which is extremely fickle and context-sensitive, I may not even confirm my hypotheses or theories before they have to be thrown out.
What’s the value of empiricism then, if it’s not fast enough to keep up with the pace of technological change? One argument might be that the objectivity that empiricism provides has a slight edge on our best intuitions, even in the short term. If I’m trying to decide what to name a particular field on a web form, am I better off with my personal biases, or small sample of data, however small or skewed? Probably with data. At least then, the design choice is defensible.
But I think there’s a better reason. What underlies these tradeoffs is the fact that many “soft” sciences are used more directly for design than the harder sciences. In all design, there is never enough data to provide an objective recommendation to every design decision. At best, the research might offer a clean delineation of the tradeoffs involved in general categories of decisions, but ultimately, it is the designer who must make the final decision, and they must make that decision with their intuition. When I design tools and notations, I might make a hundred design decisions a day, with only one backed up by empirical data.
Empiricism then, is best suited at arming the designer with the most objective and reliable intuition that data can provide. And that means that the most important part of empirical research to get right in the soft sciences is to have the designer get the data. That’s right: your programmers must observe the phenomena they will support. They must become experts in the domain. They must understand it with such detail that when faced with the smallest of programming decisions, they have an empirically grounded intuition about what will work and what will not, and a deep sense of the tradeoffs of their choice.
If we really need empiricism to cultivate intuition, what is the value of reporting empirical data? I don’t personally believe that reading a report of another researcher’s observations is nearly as enlightening. But what it can do is change what researchers look for and how they interpret their own observations. The role of the research community is to temper individual observations with a broader collection of data. This is how we generalize and validate our experiences, in the search for truth. We just have to make sure that in the process of seeking truth, we expose the designers and engineers who will be building our world to as much of reality as we can.