Dr. Lance Eliot, AI Insider
Is automation and in particular AI leading us toward a service society that depersonalizes us?
Some pundits say yes, arguing that the human touch of providing services is becoming scarcer and scarcer, and eventually we’ll all be getting served by AI systems that treat us humans as though we are non-human. More and more we’ll see and experience that humans will lose their jobs to AI and be replaced by automation that is less costly, and notably less caring, eschewing us as individuals and abandoning personalized service. Those uncaring and heartless AI systems will simply stamp out whatever service is being sought by a human and there won’t be any soul in it, there won’t be any spark of humanity, it will be push-button automata only.
In my view, those pundits are seeing the glass as only half empty. They seem to not either notice or want to observe that the glass is also half full.
Deciding On Whether There Is Depersonalization Or Personalization
We always tend to assume that whenever something is “depersonalized” that it must be bad. The word has a connotation that suggests something untoward. Nobody wants to be depersonalized.
In the case of banking ATM’s, their advent brought forth a concern that we’d no longer have the “personalized” service associated with a human bank teller, but you could argue that in some ways the ATM offers as much or even more so “personalized” service than some tellers.
The other day, I went into my local bank branch and tried to simply make a deposit via the use of a human teller. The teller tried to engage in small talk and though this was a seeming sincere attempt at personalization, in this instance I was in a hurry, had made it plainly clear that I was in a hurry, but the teller opted anyway to elongate the transaction anyway. The next day, I had another deposit that I had to make, so I went to the ATM this time, which “knew” via my prior history that I tended to make deposits and so instantly offered that option, and even indicated that it would not produce a receipt unless I requested one, since it detected that I normally don’t get a receipt.
Given the other kinds of more sophisticated patterns in my banking behavior that could be found by using AI capabilities, I thought that this ATM experience was illustrative of how even simple automation can provide a personalized service experience. Imagine what more could be done if we added some hefty Machine Learning or Deep Learning into this.
I’ve used the case of the banking effort to help illuminate the notion of what constitutes personalization versus depersonalization. Many seem to assume that if you remove the human service provider, you are axiomatically creating a depersonalized service. I don’t agree.
Take a look at Figure 1.
As shown, the performance of a service act consists of the service provider and the receiver of the service, the customer. Generally, when considering depersonalized service, we think about the service provider as being perfunctory, dry, lacking in emotion, unfeeling, aloof, and otherwise without an expression of caring for the customer. We also then think about the receiver of the service, the customer, and their reaction of presumably becoming upset at the lack of empathy to their plight as they are trying to obtain or consume the service.
I argue that the service provider can provide a personalized or depersonalized service, either one, even if it is a human providing the service. The mere act of having a human provide a service does not make it personalized. I’m sure you’ve encountered humans that treated you as though you were inconsequential, as though you were on an assembly line, and they had very little if any personalization, likely bordering on or perhaps fully enmeshed into depersonalization.
Depends On How The Automation Is Devised
It all depends upon how the automation has been established. In my view, if you add AI to providing a service, and do it well, you are going to have a solid chance of making that service personalized. This won’t happen by chance alone. In fact, by chance alone, you are probably going to have AI service that seems depersonalized.
We might at first assume that the automation is going to be providing a depersonalized service, likewise we might at first assume that a human will provide a personalized service. That’s our usual bias. Either of those assumptions can be upended.
AI Self-Driving Cars As An Example
What does this have to do with AI self-driving driverless autonomous cars?
At the Cybernetic AI Self-Driving Car Institute, we are developing AI software for self-driving cars. There are numerous ways in which the AI can either come across as personalized or depersonalized, and it is important for auto makers and tech firms to realize this and devise their AI systems accordingly.
Allow me to elaborate.
I’d like to first clarify and introduce the notion that there are varying levels of AI self-driving cars. The topmost level is considered Level 5. A Level 5 self-driving car is one that is being driven by the AI and there is no human driver involved.
For self-driving cars less than a Level 5, there must be a human driver present in the car. The human driver is currently considered the responsible party for the acts of the car. The AI and the human driver are co-sharing the driving task.
Another key aspect of AI self-driving cars is that they will be driving on our roadways in the midst of human driven cars too.
Returning to the topic of depersonalization and personalization, let’s consider how AI self-driving cars can get involved in and perhaps mired into these facets.
Bike Riders And AI Self-Driving Cars
I was speaking at a recent conference on AI self-driving cars and during the Q&A there was an interesting question or point made by an audience member. The audience member stood-up and derided human drivers that often cut-off bike riders.
She indicated that to get to the conference, she had ridden her bike, which she also rides when going to work (this event was in the Silicon Valley, where bike riding for getting to work is relatively popular). While riding to the convention, she had narrowly gotten hit at an intersection when a car took a right turn and seemed to have little regard for her presence as she rode in the bike lane.
You might assume that the car driver was not aware that she had been in the bike lane and therefore mistakenly cut her off. If that was the case, her point could be that an AI self-driving car would presumably not make that same kind of human error. The AI sensors would hopefully detect a bike rider and then appropriately the AI action planner would attempt to avoid cutting off the bike rider.
It seemed though that she believed the human driver did see her. The act of cutting her off was actually deliberate. The driver was apparently of a mind that the car had higher priority over the bike rider, and thus the car could dictate what was going to happen, namely cut-off the bike rider so that the car could proceed to make the right turn. I’m sure we’ve all had situations of a car driver that wanted to demand the right-of-way and figured that a multi-ton car has more heft to decide the matter than does a fragile human on a bicycle.
What would an AI self-driving car do?
Right now, assuming that the AI sensors detected the bike rider, and assuming that the virtual world model was updated with the path of the bike rider, and assuming that the AI action planner portion of the system was able to anticipate a potential collision, presumably the AI would opt to brake and allow the bike rider to proceed.
We must also consider the traffic situation at the time, since we don’t know what else might have been happening. Suppose a car was on the tail of the AI self-driving car and there was a risk that if the AI self-driving car abruptly halted, allowing the bike rider to proceed, the car behind the AI self-driving car might smack into the rear of the AI self-driving car. In that case, perhaps the risk of being hit from behind might lead the AI to determine that the risk of cutting off the bike rider is less overall and therefore proceed to cut-off the bike rider.
I mention this nuance about the AI self-driving car and its choice of what to do because of the oft times assumption by many that an AI self-driving car is always going to do “the right thing” in terms of making car driving decisions. In essence, people often tell me about situations of driving that they assume an AI system would “not make the same mistake” that a human made, and yet this assumption is often in a vacuum. Without knowing the context of the driving situation, how can we really say what the “right thing” was to do.
In any case, you might argue that the question brought up by the audience member is related to personalization and depersonalization. If the human driver was considering the human bike rider in a depersonalized way, they might have made the cut-off decision without any sense of humanity being involved.
Here’s what might have been taking place. That’s not a human on that bicycle, it is instead a thing on an object that is moving into my path and getting in my way of making that right turn, the driver might have been thinking. Furthermore, the driver might have been contemplating this: I am a human and my needs are important, and I need to make that right turn to proceed along smoothly and not be slowed down. The human driver objectifies the bike rider. The bike is an impediment. The human on the bike is meshed into the object.
Now, we don’t know that’s what the human driver was contemplating, but it is somewhat likely. It is easy when driving a car to fall into the mental trap that you are essentially in a video game. Around you are these various objects that are there to get in your way. Using your video gaming skills, you navigate in and around those objects.
If this seems farfetched, you might consider the emergence of road rage. People driving a car will at times become emboldened while in the driver’s seat. They are in command of a vehicle that can determine life-or-death of others. This can inflate their sense of self-importance. They can become irked by other drivers and by pedestrians and decide to take this out on those around them.
Encountering Humans In A Myriad of Ways
Back to the bike rider that got cut-off, there is a possibility that the human driver depersonalized the bike rider. This once again illustrates that humans are not necessarily going to provide or undertake personalized acts in what they do.
An AI self-driving car might or might not be undertaking a more personalized approach, depending upon how the AI has been designed, developed, and fielded.
Take a look at Figure 2.
As shown, an AI self-driving car is going to encounter humans in a variety of ways. There will be human passengers inside the AI self-driving car. There will be pedestrians outside of the AI self-driving car and that the AI self-driving car comes across. There will be human drivers in other cars, of which the AI self-driving car will encounter while driving on the roadways. There will be human bike riders, along with other humans on motorcycles, scooters, and so on.
If you buy into the notion that the AI is by necessity a depersonalizing mechanism, meaning that in comparison to human drivers the AI driver will be acting toward humans in a depersonalized manner, more so than presumably other human drivers would, this seems to spell possible disaster for humans. Are all of these humans that might be encountered going to be treated as mere objects and not as humans?
The counter-argument is that the AI can be embodied with a form of personalization that would enhance the AI driver over the at-times depersonalizing human driver. The AI system might have a calculus that assesses the value of the bike as based on the human riding the bike. Unlike the human driver of earlier mention, presumably the AI is going to take into account that a human is riding the bike.
In the case of interacting with human passengers, there is a possibility of having the AI make use of sophisticated Natural Language Processing (NLP) and socio-behavioral conversational computing. In some ways, this could be done such that the personalization of interaction is on par with a typical human driver, perhaps even better so.
Have you been in a cab or taxi whereby the human driver was lacking in conversational ability, and unable to respond when you asked where’s a good place to eat in this town? Or, the opposite extreme, you’ve been in a ridesharing car and the human driver was trying to be overly responsive by chattering the entire time, along with quizzing you about who you are, where you work, what you do. That’s akin to my bank teller example earlier.
Goldilocks Approach Is Best
AI developers ought to be aiming for the Goldilocks version of interaction with human passengers. Not too little of conversation, and not too much. On some occasions, the human passenger will just want to say where they wish to go and not want any further discussion. In other cases, the human passenger might be seeking a vigorous dialogue. One size does not fit all.
In terms of interacting with humans that are outside of the AI self-driving car, there is definitely a bit of a problem on that end of things.
Just the other day, I drove up to a four-way stop. There was another car already stopped, sitting on the other side of the intersection, and apparently waiting. I wasn’t sure why the other driver wasn’t moving forward. They had the right-of-way. Were they worried that I wasn’t going to come to a stop? Maybe they feared that I was going to run thru the stop signs and so they were waiting to make sure I came to a stop.
Well, after fully coming to a stop, I watched to see what the other car was going to do. Still no movement. I realize that most drivers in my shoes would zoom ahead, figuring that whatever the issue was about the other driver, it didn’t matter. I was concerned that the other driver might suddenly lurch forward and maybe ram into me as I drove though the intersection. They were already doing something weird, and so in my mind they were prone to weirdness of driving action.
I rolled down my window and waved my arm at the other car, suggesting that they were free to move ahead. The other driver rolled down their window, popped their head out, and yelled something unintelligible (I was too far away to hear them), and proceeded to drive forward. After the car cleared the intersection, I also proceeded forward.
In this case, we humans communicated directly with each other, albeit somewhat imperfectly. I’ve described this in my writings and speeches as the “head nod” problem of AI self-driving cars.
How will an AI self-driving car communicate with humans that are outside of the car? They cannot nod their head or wave their arms, unless we decide to put robots into the driver’s seat. Some auto maker and tech firms are outfitting the exterior of their AI self-driving cars with special screens or displays, allowing the AI to communicate via those means with humans outside of the self-driving car.
If an AI self-driving car has no means to do head nods or hand waving, it would likely be ascribed as depersonalizing that aspect of the driving act. The inclusion of special exterior screens or displays is an attempt to personalize the AI, making it seem less aloof and less non-human.
How far should we go in this? There are some concept cars that have large eyeball-like globes on the front of the self-driving car, and animated eyes that move back-and-forth, in the manner that a human eye glaze might move. Useful? Creepie? Time will tell.
Human car drivers are supposed to use their blinkers to signify their driving actions. AI self-driving cars also use blinkers. In that manner, they are the same.
Human drivers often use little micro-movements of the car, such as the tire positions and where they lean the car toward, in order to suggest driving actions that are imminent. We don’t yet have AI self-driving cars mimicking this behavior, though I’ve predicted that we ought to and will do so.
AI self-driving cars are going to be one galvanizing lighting rod of qualms about depersonalization. For the design and development of AI self-driving cars, regrettably the personalization aspects are not especially yet being given their due by many of the auto makers and tech firms. The belief by some is that those are edge or corner cases, meaning that we can wait to deal with those aspects.
Human-to-human involves personalization and depersonalization. AI-to-human also involves personalization and depersonalization. AI developers would be wise to seek the personalization side of things and overcome or avoid the depersonalization side of things.
That’s what I personally say.
For free podcast of this story, visit: http://ai-selfdriving-cars.libsyn.com/website
The podcasts are also available on Spotify, iTunes, iHeartRadio, etc.
More info about AI self-driving cars, see: www.ai-selfdriving-cars.guru
To follow Lance Eliot on Twitter: @LanceEliot
For my Forbes.com blog, see: https://forbes.com/sites/lanceeliot/
For my Medium blog, see: https://firstname.lastname@example.org
Copyright © 2019 Dr. Lance B. Eliot