Another study on MTurk found that people are OK having robots do rational tasks but not emotional tasks.
In a series of experiments, (Harvard Business School professors) Norton and Waytz tested exactly how these emotions work, and particularly whether the type of work being performed — described as either “thinking” or “feeling” jobs — changed how a person felt about a robot taking their job. The researchers turned to members of Amazon’s Mechanical Turk, the online task marketplace, to help answer these questions. Because jobs posted on the platform can vary from “evaluating advertising materials to generating a list of Bach compositions,” there’s a built-in range when it comes to emotional and cognitive tasks. The researchers then bolstered this range by presenting tasks to users as either “requiring the capacity for thinking and the ability to be rational” or requiring “the capacity for feeling and the ability to be emotional.” They then asked how comfortable they would be with a robot taking a task in general, and then specifically about taking a task they were originally supposed to perform.
“We were surprised by how consistent our results were,” reports Norton. “Again and again, we saw that while people were somewhat comfortable with robots performing ‘thinking’ types of jobs (like accounting), they became much more opposed when we asked them about robots performing ‘feeling’ jobs (like elementary school teachers).”
This is an interesting study for a couple of reasons. First, the ‘consistency in results’ is good news for validity on MTurk. Second—it makes me wonder if framing HITs as more ’emotionally oriented’ would result in more engagement from Workers (not that I have concerns with this, but with all the blah blah blah on response quality, I think this is an interesting aspect to consider).