[E]ven the most advanced machines, using the most objective measurements, can only infer how well a cup of coffee is brewed—not how it actually tastes. This is because, as a paper in the peer-reviewed journal Food Science and Management points out, the “chemistry of coffee flavor is highly complex and is still not completely understood.” It’s hard to measure what isn’t known, and coffee is estimated to contain a thousand aroma compounds. Even what can be objectively measured about a cup of coffee, its extraction and strength, “cannot tell you how good the coffee is…you do that by taste,” Vincent Fedele, whose MoJoToGo tools are widely used in the coffee industry, wrote in an e-mail. Also, there are situations where “the numbers look right but the cup can often be less than ideal.”

Last month, Christopher Mims profiled Briggio’s robot coffee kiosk, which company founder Kevin Nater says “is in essence a small food factory that absolutely replicates what a champion barista does.” Will Oremus flagged a problem with it:

Robots may be more reliable than humans, in the sense that they can work around the clock without a break and achieve levels of precision and consistency that no Starbucks employee can match. But when something goes wrong, robotic systems tend to be less resilient than those that include humans, because humans are far better at reacting to novel circumstances—not to mention soothing the feelings of unsatisfied customers.

Researchers are working on ways to allow machines to detect human emotions, but empathy is one of those human traits that is not easily automated. In general, as I’ve argued before, robots come across as clumsy and incompetent when asked to operate autonomously in human environments. That’s why the conventional wisdom is that robots are best used for work that is “dangerous, dull, and dirty, ”—work, in other words, that humans can’t or don’t want to do. The happy corollary to this is that no one complains about sewer robots or bomb-disposal robots stealing people’s jobs.