Algorithm (and RFID) Aversion

By Mark Roberti

People often do not trust technology to give them the right information or answers—even when the technology is clearly better than humans.

I recently came across some research about something called "algorithm aversion," which was carried out by academics at The Wharton School, the famed business school at the University of Pennsylvania ("Algorithm Aversion: People Erroneously Avoid Algorithms after Seeing Them Err"). The researchers found that "evidence-based algorithms more accurately predict the future than do human forecasters," yet forecasters often choose to use less accurate human forecasters. What's more, the team found that people more quickly lose confidence in algorithmic than human forecasters after seeing them make the same mistake.

This interests me, as I am often asked this question when I speak about RFID: "How do you know if the RFID system missed a tag?" That's a question I always find odd, because we clearly don't know if a human makes a mistake when counting.

There's a lack of trust in RFID. Many retailers, when running a pilot, perform manual inventory counts to determine whether the RFID count is correct. In virtually all cases, the RFID count is more accurate than hand-counting, but there is usually a senior manager who insists the RFID count must be off. He or she asks the team to carry out more manual counts until RFID is proven to be more accurate.

In my view, algorithm aversion and RFID skepticism are related. People are generally skeptical of new technologies. When bar-code systems were introduced in the United States in the early 1970s, some supermarkets put up signs assuring customers that if any bar-code scans were wrong, they would get the incorrectly scanned items for free. This was done to encourage faith in the newfangled system that was replacing the manual punching-in of the price by cashiers.

These days, faith in bar-code systems is high. Many executives seem to think that people picking up price tickets and scanning bar codes is a more accurate way to count than waving a passive UHF RFID reader around. It's not.

Back in 2009, we held an event focused on RFID in retail and apparel at the Fashion Institute of Technology, in New York City. We tagged 40 items with unique bar codes and 40 others with RFID tags, then invited one member of the audience to come up and scan the bar codes and another to read the RFID tags. The demonstration, which we videotaped, showed that an RFID tag can be missed, yes—but that a human is far more fallible.

During our demo, the audience member assigned the RFID task scanned 39 tags within 15 seconds. She missed a single tag, which we later determined was damaged when it was applied to the hangtag. But the person scanning the bar codes took three minutes and managed to count only 34 of the 40 items. With training, perhaps workers might do better than that, but the more bar codes workers scanned, the more fatigued they become, and accuracy thus drops over time. Manual counting is not only 12 times slower than RFID—or more—it's also far less accurate.

I believe this natural human inclination to be skeptical of new technologies is one reason RFID has caught on more slowly than many predicted it would. Over time, evidence does change human perception. At least, it has in the case of bar codes, which are now perceived by supermarket customers to be highly accurate. So I believe people will eventually come to trust RFID data. However, I can't say whether they will ever trust those sly algorithms.

Mark Roberti is the founder and editor of RFID Journal. If you would like to comment on this article, click on the link below. To read more of Mark's opinions, visit the RFID Journal Blog, the Editor's Note archive or RFID Connect.