Toward the search for the perfect blade runner: a large-scale, international assessment of a test that screens for “humanness sensitivity”

Abstract

We introduce a construct called “humanness sensitivity,” which we define as the ability to recognize uniquely human characteristics. To evaluate the construct, we used a “concurrent study design” to conduct an internet-based study with a convenience sample of 42,063 people from 88 countries (52.4% from the U.S. and Canada).We sought to determine to what extent people could identify subtle characteristics of human behavior, thinking, emotions, and social relationships which currently distinguish humans from non-human entities such as bots. Many people were surprisingly poor at this task, even when asked simple questions about human relationships or anatomy. Participants were best at identifying subtle aspects of human cognition and worst at identifying subtle aspects of human communication. Test scores were good predictors of whether someone was employed and modest predictors of other self-reported criterion measures. We also found that people identifying themselves in marginal societal categories (e.g., in the “other” category for gender or sexual orientation) identified themselves as less human and also scored lower on our test. As computers continue to become more human-like, our study suggests that the vast majority of humankind will likely have great difficulty distinguishing them from people. Can methods be devised for improving this ability? Might humanness sensitivity help people to make such distinctions? Will people who excel at differentiating humans and non-human entities—like the “blade runners” in the 1982 and 2017 feature films—someday hold a special place in society?

Appendices

Appendix

Debriefing

What this test is all about (seriously): nonhumans are likely to have great difficulty answering questions about unique human characteristics: our informality, idiosyncrasies, and individual styles, for example. Even more difficult for a nonhuman to fathom: extremely subtle aspects of human relationships and emotions, as well as how these and other human phenomena change as we get older. Humans also make predictable errors; when computer programs are written that imitate people, they always incorporate a serious dose of “artificial stupidity”—spelling, arithmetic, and reasoning errors, for example. To be human is to err.

Humans have bizarre dreams and daydreams. We have food cravings, especially when pregnant. We laugh at funny jokes but also when someone slips on a banana peel (what’s funny about that?). Music, art, literature and even sports sometimes make us giddy. Our memories change over time; properly designed computer memories do not. Many of us are propelled through life in a quest for money, sex, or power, or, in some cases, the perfect cup of coffee. When deeply in love, we are sometimes completely insane. We seek happiness, but some of us remain deeply depressed for months or years no matter what we do, sort of like Marvin the robot in The Hitchhiker’s Guide to the Galaxy—except that we’re real.

Many of us stubbornly believe in God or the supernatural, no matter what the facts. We sometimes become needy or whiny when sick or injured, and we feel profoundly embarrassed if we fart at the wrong time. We tell lies—but only when it’s “absolutely convenient,” as the British comedian Benny Hill put it. We divide the world into good and evil forces—both of which see themselves as good—and we sometimes commit crimes. We get headaches and tummy aches, and our hearts sometimes race when we spot an old lover. Some of us, sometimes, get tipsy or even drunk, and many of us deliberately alter our usual states of consciousness with just about any drug we can get our hands on.

It is difficult to imagine an alien, robot, or computer being able to answer any but the most trivial questions about such matters. To answer the tough questions about humans, one needs to be one.

Even with that advantage, however, most people score well under 100% on this test, mainly because, among our other foibles, not all of us understand the nuances of human relationships, emotions, defects, and idiosyncrasies. In that sense—if humanness sensitivity can be defined as “the ability to recognize uniquely human characteristics”—some of us are more human than others. There is good news, however: for the time being, even the least human of us is still more human than the most human computer.

This test was created by Dr. Robert Epstein, a distinguished research psychologist, the founder of the Cambridge Center for Behavioral Studies, and the co-creator and first director of the Loebner Prize Competition in Artificial Intelligence, an annual Turing Test in which judges try to distinguish humans from computers, first held at The Computer Museum in Boston, Massachusetts in 1991. You can learn more about the Turing Test in Dr. Epstein’s book, Parsing the Turing Test: Methodological and Philosophical Issues in the Quest for the Thinking Computer. You can download his article, “My Date with a Robot” (from Scientific American Mind) here: https://DrRobertEpstein.com/pdf/Epstein-My_Date_With_a_Robot-Scientific_American_Mnd-2006.pdf. To view Dr. Epstein interacting with a beautiful Japanese android, click here: https://www.youtube.com/watch?v=MY8-sJS0W1I

Test results are being used in an ongoing study of “humanness sensitivity” being conducted by the American Institute for Behavioral Research and Technology (AIBRT) in Vista, California, USA. AIBRT is a nonpartisan, nonprofit, 501(c)(3) research organization dedicated to improving human life. The study was approved by AIBRT’s federally registered Institutional Review Board. If you have any questions about the study or would like to be informed about the outcome of the study after the results are made public, please contact us at tests@aibrt.org. Include “EHI study” in your subject line.