If you have been attentive to artificial intelligence, you have probably seen the name Alex Pollock cropping up repeatedly. Among the louder voices warning against allowing code run wild without our signature thumbprint—our humanity—is him. He is calling into the AI lab echo chamber and grabbing the attention of the researchers, not only speaking to the choir.

Pollock’s ideas go beyond mere intellectual curiosity. According to him, human-centered artificial intelligence shouldn’t be considered as a last-minute add-on or afterglow like putting googly eyes on a Roomba and labeling it pleasant. Rather, he argues it must be weaved through design, development, and implementation from the center of the project.

Let’s break down what he means. Too many artificial intelligence systems are whirled up alone, heads-down designing away with “good enough” training data. Pollock, however, alludes to the notorious US criminal sentencing tool, COMPAS. Black defendants were disproportionately flagged as more dangerous. Why is that? thus the data was not examined under human eyes. That is a barn fire, not only a flaw.

The worst part is Pollock wants artificial intelligence teams to see beyond spreadsheet statistics. He is all about systems that make sense for actual people. Consider digital assistants learning sarcasm or translation tools picking up local language. Not every artificial intelligence today can distinguish between “I’m fine” (smoked) and “I’m fine” (with clinched teeth). They must really get us if society is to trust these technology.

One narrative Pollock enjoys telling: he collaborated with a team of healthcare artificial intelligence designers creating an elderly patient tool. The interface was first elegant but absolutely baffling. They missed talks with their real customers, seniors chafing at small text and secret menus. After a few cups of tea with their audience, changes made it straightforward, bold, and friendly. Consumption surged. The lesson here is Clever coding is subordinated to lived experience.