Dogs relish chasing after all kinds of objects such as weathered tennis ballsand worn-out frisbees.

Unfortunately, dogs also delight in chasing after passing cars, which is areally bad idea.

If you’ve ever witnessed first-hand as a brazen dog wildly ran to catch a carcoming swiftly down your neighborhood street, it is a quite scary affair andwill undoubtedly make your heartbeat uncontrollably. One possibility is thatthe driver decides to swerve or hit the brakes, any of which might end up withthe dog getting hit by the car. More likely, and usually with a more adverseoutcome, some other upcoming car will not see the dog that is running in thestreet and will accidentally ram smackdab into or entirely run over thecanine.

We know that these pell-mell pursuits of a moving car are fraught with gravedanger.

Dogs might not realize the peril and instead perceive the car as a shinyobject that is well-worth chasing. In some ways, it seems like a fun game forthe dog to be playing, at least that’s what the dog is possibly thinking.Those brash dogs that undertake this form of a daredevil sporting endeavor areusually trying to catch up with the moving vehicle, and at least are notrunning face-forward into the path of the car per se.

If the car keeps going on its existing traversal, ostensibly away from thepursuing pooch, the odds are that the dog won’t reach the vehicle and will byhappenstance escape the particular danger of getting entangled with the car.Still, there is the frightening aspect that the animal is likely now strandedin the middle of the street and other cars could fail to spy the belovedhound.

Some believe that dogs chase after cars as part of a herding instinct. Maybein the minds-eye of a dog, the car is akin to a ferocious wolf and the dog issaving all the rest of us from the wickedness of a vicious wolf-pack styleattack. Another theory is that the dog is merely utilizing its innate abilityto stalk after prey. This seems somewhat questionable since the dog istypically making all sorts of howling noises and not slinking quietly up uponthe passing car, but it nonetheless could be a strategy used for capturingquarry that potentially works in the wild and therefore might work on everydaystreets.

It is hard to know what the dog is thinking.

Perhaps there is fear involved. The dog might be intrinsically fearful of thislarge-sized unknown object that is steadily moving along. A reaction to thefear could be to lash out at the massive thing and try to stop it or maybedamage it. Some say this shows how gutsy dogs can sometimes be. This samepossibility relates to the dog as a heroic figure, trying to save thepedestrians and other humans walking around from the mysterious contraptionthat seemingly endangers everyone in the vicinity.

Trying to make somewhat lighter remarks, some jokingly suggest the dog wantsto catch and keep the car as a prized possession. Imagine how impressed theother dogs in the community would be if the capturing dog was able to showcasethat they had wrangled a car to the ground and dragged it back to theirprivate doghouse. One supposes that the proud winning dog might even dig alarge hole and try to bury the car, as though the vehicle was the largest bonethat the dog had ever managed to collect.

You might go so far as to suggest that the dog wants to drive the car (see myanalysis about dogs being trained to drive, at this link here).

Some say the real reason a dog chases after a car is because it can.

Let’s dig into that remark a bit more so.

Assuming that a human owner is responsible for the dog, that person isallowing the pooch to do something it shouldn’t be doing. In essence, it isthe fault of the human and not particularly the fault of the dog. People oughtto take care of their dogs and either keep their pets from being freely ableto chase after a car or as a minimum train their dogs to decidedly not partakein such a hazardous practice, goes this logic.

Sure, it can be challenging to teach a dog to not chase after cars. Thearduous task can involve hours upon hours of careful and conscientious effortwith your treasured pet. There isn’t much argument about the difficultiesinvolved. That being said, the time is well spent and necessary for the well-being of all parties.

The other obvious point to be made is that if a dog is routinely kept on aleash, for which most locales have pointed leash laws, there is not muchchance that a dog can indeed chase after a car. Unless the human relinquishesthe leash or allows their dog out of a fenced area, presumably the opportunityto actively give chase after a car is nearly nonexistent.

Of course, slipups can occur. The most faithful of owners could inadvertentlylet the leash escape from their grasp while walking their dog, especially ifthe enthused animal suddenly and surprisingly darts away without any warning.Or perhaps a fenced area has a gate that wasn’t properly locked, and the dogshoved at the potential opening while fiercely intent on watching that cartravel down the street. For those myriads of reasons, training a dog to notchase a car has added merits.

Why all this talk about dogs chasing cars?

Consider that the future of cars consists of self-driving cars. In the case oftrue self-driving cars, there won’t be a human driver at the wheel. An AIdriving system will be in charge of the driving controls and determining howthe car will be driven (for more details about self-driving cars, see mydiscussion at this link here).

Tying together these two topics, it will be important for AI driving systemsand thus self-driving cars to be able to contend with dogs that opt to chasecars. Yes, it would seem just as likely that a dog might chase a self-drivingcar as they would a human-driven car.

Today’s intriguing question is this: How will an AI-based true self-drivingcar contend with dogs that chase cars and can we rest easy knowing that thoseself-driving cars will do the right thing?

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 carentirely on its own and there isn’t any human assistance during the drivingtask.

These driverless vehicles are considered Level 4 and Level 5 (see myexplanation at this link here), while a car that requires a human driver toco-share the driving effort is usually considered at Level 2 or Level 3. Thecars 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 asADAS (Advanced Driver-Assistance Systems).

There is not yet a true self-driving car at Level 5, which we don’t yet evenknow if this will be possible to achieve, and nor how long it will take to getthere.

Meanwhile, the Level 4 efforts are gradually trying to get some traction byundergoing very narrow and selective public roadway trials, though there iscontroversy over whether this testing should be allowed per se (we are alllife-or-death guinea pigs in an experiment taking place on our highways andbyways, some contend, see my coverage at this link here).

Since semi-autonomous cars require a human driver, the adoption of those typesof cars won’t be markedly different than driving conventional vehicles, sothere’s not much new per se to cover about them on this topic (though, asyou’ll see in a moment, the points next made are generally applicable).

For semi-autonomous cars, it is important that the public needs to beforewarned about a disturbing aspect that’s been arising lately, namely thatdespite those human drivers that keep posting videos of themselves fallingasleep at the wheel of a Level 2 or Level 3 car, we all need to avoid beingmisled into believing that the driver can take away their attention from thedriving 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 Those Dogs That Chase Cars

For Level 4 and Level 5 true self-driving vehicles, there won’t be a humandriver involved in the driving task.

All occupants will be passengers.

The AI is doing the driving.

One aspect to immediately discuss entails the fact that today’s AI is notsentient. In other words, the AI is altogether a collective of computer-basedprogramming and algorithms, and most assuredly not able to reason in the samemanner that humans can. For example, current AI lacks any semblance of commonsense and nor has any kind of common-sense reasoning (for further background,see the link here).

This is important to keep in mind when considering the matter of a dog thatchases after a self-driving car.

A human driver would realize the significance of a chasing dog. The driverwould comprehend that a dog is, well, a dog. It can scamper and run. It can dothings that aren’t wise to do. It can veer this way and that way, nearlyerratically. The dog might pursue the chasing car and abruptly decide it hashad enough and come to a halt. Or the dog might run up onto lawns and acrosshomes to try and reach the car, thus not taking a straight path towardreaching the in-motion vehicle.

None of that is “understood” within an AI driving system, and even if the AIdriving system is programmed to cope with the act of dogs that chase cars,keep in mind that the programming is only programming and does not have therobustness of human thinking intertwined.

Consider this weighty facet in a somewhat different light.

Suppose you had a teenage novice driver that was just initially in the throesof learning to drive. The neophyte is barely able to use the pedals and thesteering wheel, in the sense that they are unsure of how to perform thedriving task.

If that teenager was driving in a local neighborhood and all of a sudden a dogstarted to give chase, there is no question that the young adult wouldindubitably comprehend the significance of what a dog is, how a dog acts, andthe qualms associated with having a dog chasing their car. Unless the teenagerhas been living in a cave their whole life, they would inarguably know whatdogs are and what they can do.

I’m not suggesting that the novice driver would know what to do in such adriving circumstance and simply emphasizing that the human driver wouldrealize the nature of dogs. This goes into the earlier exhortation about asemblance of common sense. Despite those that might chortle about ascribingcommon sense to teenagers, you must admit that a teenager is quite likely tohave rather extensive knowledge and awareness about dogs.

An AI driving system of today is unlikely to have anything at all thatprovides background about dogs. There is no overarching reason to go to thetrouble to infuse this kind of information into an AI driving system at thistime (exceptions do apply). The developers have enough on their hands in termsof getting a self-driving car to be able to safely go from point A to point B,and thus the so-called edge case or corner case involving a chasing dog isvery low on the priority list.

In essence, the AI driving system might do something like this.

First, realize that the sensors of the self-driving car usually include videocameras, radar, LIDAR, ultrasonic units, and the like. These are akin to beingthe eyes and ears of the AI driving system. Via those sensors, the AI drivingsystem interprets the data being collected and attempts to discern the drivingscene. Aspects include notable particulars such as where is the roadway, arethere are other cars nearby, are there pedestrians nearby, and so on.

Trying to detect a dog is not particularly easy.

Dogs tend to be relatively low to the ground. Their movement can be quitequick. With pedestrians, you can oftentimes examine a visual image to figureout which way a pedestrian is possibly going to turn, or whether they have gottheir arms out and maybe are going to run into the street. Trying to do asimilar analysis about dogs is relatively hard to do.

So, you’ve got a dog, as an object, which is low to the ground, moving fast,and difficult to predict in terms of what the object might do.

On top of those complexities, let’s assume the dog is coming up toward theself-driving car, doing so from behind the vehicle. By-and-large, the focus ofthe sensors and the AI driving system has to do with what’s ahead of the car,not especially what is behind the car (obviously, another car that is comingup fast from behind and that might ram the self-driving car is a vitalconsideration, though realize that another vehicle is a relatively largerobject, more commonly expected, and generally easier to detect and predict).

All in all, the bad news for the dogs that chase cars are that few if anyself-driving cars of today are prepared to cope with the specifics of a dogthat is chasing the car. There are (at this time) not many AI driving systemsthat have a pre-programmed response to deal with a chasing dog. And, asmentioned, even trying to detect the dog is problematic, to begin with.

In the discussion earlier it was pointed out that the car being chased isperhaps less likely to end-up striking a dog and it might be more likely thatother nearby cars might do so, assuming that there is other traffic. In thatuse case, the odds are a little better about the AI-based self-driving carsbecause they are perhaps diligently scanning the area and trying tocontinually ferret out what is going on in the driving scene.

Human drivers oftentimes will drive while distracted, ironically watching catvideos just when a dog happens to be in the middle of the street. The hope isthat AI driving systems will be as safe or possibly safer than human drivers,partially due to the lack of distractions and also since the AI won’t bedrinking and driving.

In short, one can be somewhat hopeful that if a self-driving car comes upon asituation wherein a dog is chasing after a car that is ahead of the self-driving car, there is some semblance of a chance that the AI driving systemmight detect the dog. Though the AI might not be able to classify what theobject is, nonetheless, there is an object that has been detected and ispresumably in the roadway, thus, it becomes an object to be given dueconsideration.

Most AI driving systems would interpret the object as something to be trackedand monitored, waiting to ascertain what the detected object is going to do.From the AI programming viewpoint, the object might just as well be atumbleweed that has come onto the street and is weaving to and fro. Of course,a car can ram into a tumbleweed if needed, but ramming into a dog or otherliving creature is a different matter (see my discussion about roadwayintervening animals and self-driving cars at this link here).

We can add a variety of twists to this dilemma.

A human driver is driving along, and a dog starts to chase after the movingcar. A self-driving car is behind the car that is being chased by the dog. Oneconcern involves the dog. Another concern involves the potential actions ofthe human driver.

Envision that the human driver in this scenario decides to “freak out” andslams on their brakes. Maybe the person believes that the best course ofaction is to come to a sudden halt. The dog will then be able to catch-up withthe car. The human driver doesn’t have to worry about running over the dogsince the car is now at a standstill.

Seems like impeccable logic.

Don’t though forget that there is another car immersed in this drivingscenario, namely the self-driving car that is behind the human-driven car. Itis conceivable that the abrupt act of the human driver might lead to the self-driving car rear-ending the human-driven car. This could certainly happen tooif the car that was behind the stopped car was being driven by a person, andthus you cannot somehow ascribe that this would only happen to an AI-basedself-driving car.

The crux of this twist is that the AI driving system has to be programmed toanticipate the reactions of other objects in the driving scene, including whatthe human-driven cars are going to do or react in response to the bolting dog.

Another object that might get involved in this expanding scenario could be apedestrian or perhaps a slew of pedestrians, all of whom suddenly decide theywill come into the street to try and save the dog. You can easily imagine thiskind of reaction. People that see a dog in danger are likely to take dangerousactions themselves, doing so in hopes of saving the dog and withoutnecessarily giving due thought to their own safety, and nor the safety of thepassengers in the cars entangled in the entire affair.

The point is that a dog chasing a car can be an entangling web of activitiesthat rapidly unfold or evolve while the driving action is underway. If youonly construe the matter as one dog and one car, you have demonstrablyunderstated the nature of the problem and the life-or-death consequences thatcan arise.


Some believe that dogs won’t chase after a self-driving car.

Here’s why they say this.

Dogs are aware today that a human is at the wheel, presumably being able tosee the person, and therefore they are chasing not merely a car, they are alsochasing the person. Since an AI-based true self-driving car does not have ahuman driver at the wheel, dogs will no longer feel any need or incentive tochase after cars.

Sorry, but as they say, that logic won’t hunt.

First, it seems questionable to assume that dogs can connect the notion that acar is being driven by a human with the aspect that this big shiny object iscoming down the street. In many cases, the human driver is not readily seenanyway. All in all, the odds are higher than the dog is chasing the car, andnot somehow trying to chase the human driver (to clarify, sure, there arerarer instances of the dog spying their owner driving a car and wanting tofollow their owner, that kind of situation).

Secondly, there are going to be passengers inside self-driving cars. Not allof the time, since it is abundantly possible for self-driving cars to beroaming around without passengers, or possibly doing a delivery chore andthere is no need for any riders inside. In any case, for likely much of thetime, there are going to be people riding inside self-driving cars.

Can the dog distinguish between humans that are driving a car versus humansmerely residing in a car?

Look, I am the first to say that dogs are darned smart, and we ought to givethem immense credit, but it seems rather farfetched to claim that dogsprincipally choose to chase cars due to there being a human driver at thewheel, and furthermore that these savvy pooches would ergo also mentallycalculate that human riders inside a self-driving car are not serving as humandrivers.

Since my dog does the balancing of my checkbook each month for me, I’ll bringthis up during the time that we have together for doing my accounting and seewhat the dog says. Probably will cost me some of his favored dog biscuits, butit will be worth it to find out exactly why dogs do indeed chase cars.

I’ll let you know.

Source: LanceEliot

Lance Eliot



Dr. Lance B. Eliot is a world-renowned expert on Artificial Intelligence (AI)and Machine Learning (ML).

Previous Raw pet food safety risks rise with owner cognitive bias

Next PFMA releases latest UK Pet Population Data