AI politeness could be inconsiderate.
Does it seem to you that civility and politeness are on the wane?
Many would surely agree that politeness appears to have been placed on the back-burner, or perhaps dropped altogether from our lexicon. No one can put their finger on the precise reason for this increasingly slippery slope of tossing politeness to the winds.
Politeness arises not only among humans, but also can appear when interacting with automation. There is a veritable tsunami of AI systems being developed and fielded, many of which make use of Natural Language Processing (NLP). You are likely familiar with Alexa and Siri, which are prime examples of today’s capabilities at NLP and how such state-of-the-art AI systems interact with humans.
At first, the use of AI NLP was undertaken in a somewhat neutral or non-polite mode.
The thought by AI developers was that the messages emitted by the AI should be direct and to the point, thus, no need to add those (assumed) superfluous aspects to make the wording perceived as being polite.
Gradually, it was realized that the starkness of the wording was a giveaway that the AI was a machine and not a human. As such, many NLP systems not only added the politeness trimmings, they went further and have added pauses as though the AI is “thinking” and included filler sounds that a human might proffer such as “uh-huh” and other common human-like expressions.
On the surface, this effort to make AI seem more conversational appears to make a lot of sense (see my detailed explanation at this link here). People will be put at ease and feel more comfortable chatting with AI. The interaction would be less clumsily disjointed. Others worry, rightfully so, that people will get a false or misleading impression of what the AI can do.
By adding the bells and whistles of human-to-human interaction into AI-to-human interaction, there is a real danger of leading people down a primrose path. You might inadvertently assume that the AI has a fully conversant and sentient capacity, along with believing that the AI has common-sense reasoning (all of these are human qualities that AI has yet to fully achieve, see my explanation at this link here).
Typically, when AI acts in this way, humans will anthropomorphize the system and fall into the mental trap that the AI is truly equal to human intelligence.
Politeness is a dual-edged sword that provides an added indicator potentially of human qualities and thus could be considered part of the anthropomorphizing tendency sway.
Using politeness in AI is darned if you do, darned if you don’t, kind of challenge.
When you are polite, it oftentimes is done as part of social interaction dynamics and can encompass power differentials. Being impolite to someone can be an obvious way to showcase one person’s greater social status over another.
Too, being polite can have a similar effect.
Consider these three somewhat vexing questions:
· Should AI be polite?
· Should AI be neutral (non-polite)?
· Should AI be impolite?
You might be tempted to say that the AI can be any of those three, at any time, in any mixture, as dependent upon the context and the culture so involved. That’s quite a mouthful and trying to discern the right level of politeness for the right moment and to the person(s) involved can be trickily problematic.
There are various AI research efforts trying to figure out these aspects, including taking non-polite human utterances and doing an automated translation into polite versions, using Machine Learning and Deep Learning, for which this is a lot harder than you might think (for an interesting study by researchers at Carnegie Mellon University on politeness transfer, see this link here).
Shifting gears, consider the various situations that we will inevitably be interacting with an AI NLP.
One such instance consists of the advent of AI-based true self-driving cars and the need for people to interact with the AI driving system.
Here’s an intriguing question: Does AI politeness bode for positive or negative aspects when humans are passengers in AI-based true self-driving cars?
Let’s unpack the matter and see.
Understanding The Levels Of Self-Driving Cars
As a clarification, true self-driving cars are ones that the AI drives the car entirely on its own and there isn’t any human assistance during the driving task.
These driverless vehicles are considered a Level 4 and Level 5 (see my explanation at this link here), while a car that requires a human driver to co-share the driving effort is usually considered at a Level 2 or Level 3. The cars that co-share the driving task are described as being semi-autonomous, and typically contain a variety of automated add-on’s that are referred to as ADAS (Advanced Driver-Assistance Systems).
There is not yet a true self-driving car at Level 5, which we don’t yet even know if this will be possible to achieve, and nor how long it will take to get there.
Meanwhile, the Level 4 efforts are gradually trying to get some traction by undergoing very narrow and selective public roadway trials, though there is controversy over whether this testing should be allowed per se (we are all life-or-death guinea pigs in an experiment taking place on our highways and byways, some point out, see my indication at this link here).
Since semi-autonomous cars require a human driver, the adoption of those types of cars won’t be markedly different than driving conventional vehicles, so there’s not much new per se to cover about them on this topic (though, as you’ll see in a moment, the points next made are generally applicable).
For semi-autonomous cars, it is important that the public needs to be forewarned about a disturbing aspect that’s been arising lately, namely that despite those human drivers that keep posting videos of themselves falling asleep at the wheel of a Level 2 or Level 3 car, we all need to avoid being misled into believing that the driver can take away their attention from the driving task while driving a semi-autonomous car.
You are the responsible party for the driving actions of the vehicle, regardless of how much automation might be tossed into a Level 2 or Level 3.
Self-Driving Cars And Politeness
For Level 4 and Level 5 true self-driving vehicles, there won’t be a human driver involved in the driving task.
All occupants will be passengers.
The AI is doing the driving.
When you are a passenger in a human-driven car, the driver might be polite, they might be non-polite, and of course, they at times might be impolite.
What should the AI driving system be like?
Imagine that you are heading to work and using a self-driving car (see this link for what it’s like to ride inside a self-driving car).
The AI that is driving the car might opt to explore behind-the-scenes your usual morning rides via data stored in the cloud by the automakers and self-driving car tech firms and detects a pattern that you usually opt to have the AI swing through a Starbucks to get your favored latte, helping to awaken you for the day’s work efforts ahead.
Rather than waiting for you to suggest the driving diversion, the AI instead talks to you via the AI NLP capability and actively prompts you about it.
Here’s what the AI might say (three variants):
· Polite: “If it pleases you, I can visit the Starbucks drive-through so you can get your morning latte.”
· Non-Polite (neutral): “Should I visit the Starbucks drive-through so you can get your morning latte?”
· Impolite: “Hey, baggy-eyed, do you want me to visit the Starbucks drive-through so you can get your morning latte?”
The polite version includes an elongated instance of the “please” aspects, which the non-polite excludes, and meanwhile the impolite throws a barb at you about being baggy-eyed (perhaps the inward-facing camera analyzed your face using facial recognition and determined that your eyes are baggy, though even if so, it seems impolite to especially make note of your morning-weary facial features).
Some people might be creeped out that the AI “knew” about their morning routine of getting a latte, and I just note that this is part of the upcoming privacy intrusion issues that will rear its ugly head as true self-driving cars become prevalent (see my analysis of those privacy issues at this link here).
Anyway, some people meanwhile will relish having a “chauffeur” type of personalized driver.
Back to the politeness topic, you might be upset and disgusted at being called baggy-eyed. It could ruin your day and cause you to render a complaint with the owner of the self-driving car.
Or, one supposes, you might be tickled by the remark, taking it as a wake-up call, as it were, and nod your head approvingly that you did stay out last night until the late hours and today seem to be paying the price.
You could even argue that the polite version has gone too far, mentioning “if it pleases you” and this might upset a rider. What pleases me is for you to shut-up, the passenger might retort.
Okay, so it is not blatantly obvious which of the utterances is best, and the AI ought to be considering contextual parameters to ascertain the politeness level and encumbrances to be used (not easy for the AI, and not easy for humans to do either).
Conclusion
There is an added twist to self-driving cars politeness as a use case.
Keep in mind that the AI is undertaking life-or-death driving actions when serving as the driving system for the self-driving car.
Suppose the AI has detected an upcoming dangerous roadway condition and wants to interact with the passenger about options available for dealing with the pending matter. Ultimately, the AI is going to decide what to do but wants input or preferences from the passenger.
This might involve urgently asking questions of the rider, any wording of which might be easily misunderstood or misinterpreted by the passenger, and thus the human in this AI-to-human interaction might provide an answer that is askew of what otherwise would have been intended.
Or, the AI might be seeking to forewarn the passenger, such as the AI is alerting the passenger to get ready for a rapid and jarring stop because a pedestrian has unexpectedly jaywalked in front of the in-motion self-driving car.
Perhaps this is what the AI might emit: “Excuse me, but I regret to inform you that I am going to need to take a rather abrupt application of the brakes, so please do get yourself readied and know that I am going to come to a stop as safely as possible, and thanks for your attention to this indubitably pressing matter.”
Some wonder whether politeness might get in the way of those crucial interactions.
Maybe this works instead: “Jamming on the brakes, hold on!”
