In January 2021, a renowned international tech company publicly announced that it would install cameras with behavioral biometric profiling technology in its transport vehicles in order to prevent accidents resulting from driver fatigue. The news was not so well received by the drivers of the distribution companies, with some immediately threatening to resign.

Now that such technology is readily available, we can see how the international market has been flooded by agents who, taking advantage of the developments in artificial intelligence, promote technological services that can detect levels of fatigue, drowsiness, distraction and recklessness while driving, offering to reduce the risk of accidents related to these factors. The technical requirements are relatively simple: A series of cameras and sensors are installed inside and outside the vehicle, allowing to capture both its movements and the driver’s; this data is subsequently interpreted according to a specific algorithm (for example, that associates eye deviations as distraction, or the slow and repeated movement of the eyelids as drowsiness).

However, what may be seen as an opportunity for companies to reduce costs associated with work accidents, liability to third parties and infrastructure damage could be perceived from the drivers’ side as undue intrusion into their privacy and could open the door to eventual continuous surveillance, subjecting workers to perpetual stress.

Despite the fact that Chile has an outdated (and, hopefully, soon to be repealed) personal data protection law, Chilean agencies responsible for the protection of workers’ rights have issued certain pronouncements regarding this matter. These pronouncements have resulted in an application of the robust guiding principles of labor law to supplement the currently limited privacy regulations. While the intention is to protect the right to privacy, is it the right approach, in light of these emerging technologies?

Surveillance of workers: Effective or excessive?

The Chilean Labor Bureau has a longstanding doctrine regarding the use of video surveillance systems in the workplace. In this regard, it has stated that it is not lawful to use such control mechanisms for the sole purpose of monitoring and supervising worker activity since it constitutes a form of unlimited control, although such surveillance is lawful when objectively necessary due to technical or manufacturing process requirements, or for safety reasons of workers or third parties (criterion adopted from 2002 onwards). Moreover, the Supreme Court issued a ruling in 2017 in which it declared unlawful the use of surveillance cameras inside urban transport buses if the camera directly captures the driver’s face, as doing so does not serve any of the abovementioned lawful purposes.

Furthermore, in January 2021, the Labor Bureau ruled on the use of artificial intelligence systems to check for fatigue, drowsiness and distraction of truck drivers of the state-owned oil company Empresa Nacional del Petróleo (ENAP), stating that the use of this technology resulted in disproportionate control that unduly affected workers’ right to privacy. Although the Bureau recognizes that the purpose of preventing accidents and protecting the life and health of the company’s workers is legitimate and could, eventually, justify the intrusion into their privacy, there are a number of factors whose confluence makes these measures disproportional, such as:

1. Requiring certain driver behaviors (such as not yawning) is contrary to nature and closer to automated actions rather than human ones;
2. The constant surveillance the driver is subjected to generates intolerable levels of alertness and stress.

The first point seems to be explained by the characteristics of the system in question. According to the report presented by ENAP, the mere act of yawning could trigger a warning consisting in the AI system telling the worker to “please take a break,” in circumstances where this act alone is not necessarily indicative of a state of drowsiness that increases the possibility of an accident. This situation is in some ways a reflection of what happened between the tech company mentioned above and its delivery drivers who accused the company of abusive practices. Chief among the drivers’ complaints was that the system mistakenly qualifies “risk events” (for example, they claim that every time they are overtaken by another vehicle, the algorithm warns that the delivery driver is not keeping a safe distance), which results in an increase of stress due to incessant criticism from a robotic voice, and also affects the workers’ qualification, negatively impacting their performance benefits. While this seems to be a mere software issue, it is in fact connected to the second factor, as explained below.

The second point may give rise to a more interesting discussion on both legal and ethical grounds. The truth is that we live in a “new normal,” in which the constant monitoring of our lives (by our own cell phones and wearables) requires us to make constant concessions regarding our privacy. In that regard, it is essential for us to know the difference between actual vigilance and the mere perception of it, which will allow us to properly analyze whether a specific scenario falls under the umbrella of “prohibited uses” dictated by Chilean institutions.

According to the reports delivered by ENAP, the cameras did not constantly record driver activity. They sensed movement and only began recording if a risk event was perceived; thus, the cameras did not function as “supervising devices.” Moreover, the recordings were never sent to the driver’s employer, but only to a third-party monitoring center. The employer never gave instructions to drivers in respect of the information recorded, nor did driver infringements negatively impact their salary or position within the company (at least according to the information made available in the Labor Bureau ruling).  Considering that we normally allow our own vehicles to alert us about safety measures (such as door locking, getting too close when parking, etc.) it doesn’t seem that ENAP’s systems are problematic by definition; moreover, it is possible that the Labor Bureau could be confusing “effective surveillance” with the “perception of being under surveillance.”

That said, the perception of hypervigilance should not be disregarded as a minor issue since it could eventually produce the same amount of stress and psychological damage as effective continuous vigilance. If this is the problem the authorities have regarding behavioral biometric profiling technology, we believe it could be addressed by training the algorithm to reduce the number of reported false positives. In this way, it would be reasonable to conclude that if the alert system were activated only when a real danger was imminent, drivers would not feel that their sphere of privacy was being so strongly violated.

Thus, these kinds of problems connected with such AI-powered operator monitoring systems may be due to a training error in the algorithm that does not allow it to properly classify data that accounts for a risk event and data that account for a false positive; likewise, it would not be surprising if the quality of the training data was deficient. This was an issue that has already cost automatic car companies in China reputational damages after some users publicly stated that the AI system in their car interpreted the natural slanted shape of the driver’s eyes as a sign of fatigue.

That said, if these technical issues can be resolved in a proper manner—and as long as employers comply with the rest of the regulatory privacy requirements—we can rechannel our Orwellian fears of technology and begin to see it as a potential ally, especially when its use can prevent tragedies for drivers and third parties. This change in perception could allow courts and other agencies to make reasonable case-by-case assessments, instead of setting general rules that view privacy as an absolute and inviolable right.