Did the robot that saved Spooner's life in I, Robot break the 2nd law of robotics?

Question

There is a particular scene in I, Robot that raises an interesting issue. Detective Spooner is telling the story of how he lost his arm. He was in a car accident, and found himself and a young girl underwater, about to drown. Will was saved by a robot that was passing by. The reason given as to why the robot saved Will and not the girl was that, statistically speaking, Spooner's chances for survival were higher than the girl's. That makes sense, as a robot would make decisions based on statistical probability. But there is a problem: In the flashback, we clearly hear Will's character shout, "Save the girl!" to the robot. That was a direct order from a human to a robot.

At first I thought that this was not a violation of the 2nd law because if the robot had obeyed his order, then he would have died. The 2nd law of robotics cannot override the 1st law. But the problem with this is that, if a Sophie's choice type situation counts as breaking the 1st law of robotics, then an emergency robot would never be able to perform triage (as the robot that saved Spooner clearly did). If choosing to save one human counts as harming another human, then a robot would not be able to effectively operate in an emergency such as this.

All this leads to my question:
Did this robot break the 2nd law of robotics, or is there some nuanced interpretation of the laws that could explain its behaviour?

Answer 1

The Second Law states

A robot must obey the orders given it by human beings except where such orders would conflict with the First Law.

So you're saying by refusing to save the girl as ordered by Detective Spooner, the robot has broken that law? The only way it can't have broken the second law is if the corollary comes into play and it would conflict with the first law.

The First Law says

A robot may not injure a human being or, through inaction, allow a human being to come to harm.

Obviously, the robot has not directly injured anyone so that is out of the picture. The robot has not allowed the girl to come to harm by inaction as the robot is acting by helping Spooner (i.e. it isn't just standing there watching). However, if, as you say, it was statistically more likely that Spooner would survive with some help, then obeying Spooner and helping the girl could be construed as letting Spooner come to harm by inaction.

So the Second Law is not broken as it's over-ruled by the First Law. The First Law is not broken as the robot did its best to save a human.

Answer 2

The film appears to operate on anachronistic Asimov mechanics

What we would have here is likely a first-law vs first-law conflict. Since the robot can not save both humans, one would have to die.

I, Robot era:

There is definitely precedent for an I, Robot era robot knowingly allowing humans to come to harm, in the short story "Little Lost Robot", but this was under the circumstance that the human would come to harm regardless of the robot's action, so the robots deem that it is not through their inaction that the humans come to harm.

However, I would suspect that instead, an Asimov robot would interpret the situation in the film as a first-law vs first-law conflict, since either human could be saved depending on the robot's decision. In other words, the robot could have saved the child, but didn't, which would be a first law violation. Looking at both victims this same way, the robot would then find this to be a first-law vs first-law conflict.

The short story "Liar" explores what happens when a robot is faced with a first-law vs first-law scenario:

Through a fault in manufacturing, a robot, RB-34 (also known as Herbie), is created that possesses telepathic abilities. While the roboticists at U.S. Robots and Mechanical Men investigate how this occurred, the robot tells them what other people are thinking. But the First Law still applies to this robot, and so it deliberately lies when necessary to avoid hurting their feelings and to make people happy, especially in terms of romance.

However, by lying, it is hurting them anyway. When it is confronted with this fact by Susan Calvin (to whom it falsely claimed her coworker was infatuated with her - a particularly painful lie), the robot experiences an insoluble logical conflict and becomes catatonic.

In short, an I, Robot era robot in Asimov's writing would not have been able to continue functioning after this scenario and would have to be discarded completely. It's likely that it would not even be able to function after being initially faced with the scenario, thereby destroying itself before being able to rescue either human.

The second law is irrelevant, because first-law vs first-law results in an unsurvivable deadlock. First law is the "trump card" so to speak, and not given a priority, lest the second or third compete, as we see in Runaround:

In 2015, Powell, Donovan and Robot SPD-13 (also known as "Speedy") are sent to Mercury to restart operations at a mining station which was abandoned ten years before.

They discover that the photo-cell banks that provide life support to the base are short on selenium and will soon fail. The nearest selenium pool is seventeen miles away, and since Speedy can withstand Mercury’s high temperatures, Donovan sends him to get it. Powell and Donovan become worried when they realize that Speedy has not returned after five hours. They use a more primitive robot to find Speedy and try to analyze what happened to it.

When they eventually find Speedy, they discover he is running in a huge circle around a selenium pool. Further, they notice that "Speedy’s gait [includes] a peculiar rolling stagger, a noticeable side-to-side lurch". When Speedy is asked to return with the selenium, he begins talking oddly ("Hot dog, let’s play games. You catch me and I catch you; no love can cut our knife in two" and quoting Gilbert and Sullivan). Speedy continues to show symptoms that, if he were human, would be interpreted as drunkenness.

Powell eventually realizes that the selenium source contains unforeseen danger to the robot. Under normal circumstances, Speedy would observe the Second Law ("a robot must obey orders"), but, because Speedy was so expensive to manufacture and "not a thing to be lightly destroyed", the Third Law ("a robot must protect its own existence") had been strengthened "so that his allergy to danger is unusually high". As the order to retrieve the selenium was casually worded with no particular emphasis, Speedy cannot decide whether to obey it (Second Law) or protect himself from danger (the strengthened Third Law). He then oscillates between positions: farther from the selenium, in which the order "outweighs" the need for self-preservation, and nearer the selenium, in which the compulsion of the third law is bigger and pushes him back. The conflicting Laws cause what is basically a feedback loop which confuses him to the point that he starts acting inebriated.

Attempts to order Speedy to return (Second Law) fail, as the conflicted positronic brain cannot accept new orders. Attempts to force Speedy to the base with oxalic acid, that can destroy it (third law) fails, it merely causes Speedy to change routes until he finds a new avoid-danger/follow-order equilibrium.

Of course, the only thing that trumps both the Second and Third Laws is the First Law of Robotics ("a robot may not...allow a human being to come to harm"). Therefore, Powell decides to risk his life by going out in the heat, hoping that the First Law will force Speedy to overcome his cognitive dissonance and save his life. The plan eventually works, and the team is able to repair the photo-cell banks.

Robot novels era:

A few thousand years after the I, Robot era, the first-law vs first-law dilemma has essentially been solved.

In The Robots of Dawn, a humaniform robot experiences a deadlock and is destroyed, and Elijah Bailey is tasked with discovering why. He suggests to Dr. Fastolfe, one of the greatest roboticists of the age as well the robot's owner and creator, that a first-law vs first-law dilemma might be responsible, citing the story of Susan Calvin and the psychic robot. However, Dr. Fastolfe explains that this is essentially impossible in the modern age because even first law invocations are given a priority and equal priorities are selected between randomly; that he himself is probably the only person alive who can orchestrate it, and it would have to be on a good day.

We see direct instances of robots handling priority in first law conflicts throughout the novels, such as in The Naked Sun, when another humaniform robot forces Bailey to sit so that it can close the top on a transporter to protect him from his agoraphobia.

The disadvantage is that it is possible, though requires extreme circumstances, for multiple second-or-third-law appeals to outweigh an appeal to the first law, as we again see in The Robots of Dawn that Bailey notices a group of robots are willing to overlook his injuries when he insists that they are not severe and casually instructs them to go about their business. He knows that this command can not outweigh the appeal to the first law, and so he reasons that the robots have been given very strict instructions in addition. The two commands and his own downplay of the severity of his situation, he reasons, raise the priority of the second law to surpass that of the first law.

The robot in question in the film is said to have decided that one human had a greater chance of survival than the other, and used that information to determine which human to save. This would not be a factor in the I, Robot era, but is a fact of basic robotics in the robot novels era. However, it would seem Spooner's command to save the girl instead is not of sufficient priority to outweigh the difference in priorities between his own first law appeal and the child's.

Answer 3

I am going to look at this question from a logical real-world point of view.

The robot does not break the second law; but technically, it does break the second. That said, the rules would only be a condensed explanation of far more intricate logic and computer code.

To quote Isaac Asimov's laws of robotics, emphasis mine:

Rule 1:
A robot may not injure a human being or, through inaction, allow a human being to come to harm.
Rule 2:
A robot must obey the orders given it by human beings except where such orders would conflict with the First Law.
Rule 3:
A robot must protect its own existence as long as such protection does not conflict with the First or Second Laws.^{Irrelevant, but provided for completions sake.}

In the given situation, the robot had to act, in order to save either Will or the child. The robot, being capable of calculating odds to a near-perfect prediction rate, is able to establish that it can not act in enough time to save both Will and the child; and the odds say that Will has a better chance to survive. In such a case, it is only logical to choose Will. This is a key plot point; robots run off pure logic - a fact that makes them differ greatly from human beings.

When the robot fails to also save the child, it is causing harm through inaction. However, again, the robot knows that it is impossible to save both Will and the child. It picks the best option, in order to best adhere to its rules. This is where a greater understanding of computers and the rules, themselves, come in to place.

What would actually happen, considering the explicit rule

The rules are not an absolute fact. They are not there to say "robots will never harm a human, and will always save the day, when present". We know this by how the movie plays out. The rules are simply the rules used to govern the actions of the robots, in-verse. As a programmer, this is something that is blatantly obvious to me; but I am confident that it is not as so for others that are not familiar with how strictly adherent any computer system is.

The point is, the rule does not state anything about it "not counting" because the robot is "already saving someone". As such, only considering this explicit rule (as any computer or robot would interpret, at least), there is no allowance for a situation where the robot can only save one of two people in danger. In actual computer science, only considering the explicit rule, such an event would likely cause an infinite loop. The robot would stop where it was, and continue to process the catch-22 forever; or at least, until its logic kicked it out of the thought process. At this point, the robot would dump its memory of the current event, and move on. In theory.

What would probably happen, in-verse

In verse, the rules are a lot more complicated; at least, internal to the robot. There would likely be a whole lot of special cases, when processing the core rules, to determine how to act in such situations. As a result, the robot is still able to act, and takes the most logical outcome. It only saves Will, but it does save someone.

It is far more understandable that the rules would be simplified to three generic common-case situations; it would be far less believable that people would be so easily trusting of robots if the rule read "A robot may not injure a human or, through inaction, allow a human being to come to harm; unless in doing so, there is greater chance of preventing another human from coming to harm". There are just way to many ways to interpret this.

---

So as far as the explicit rules go, the robot does not break the second rule; disobeying Will's action does not go against "preventing a human from coming to harm through inaction", because through disobeying Will, it saves Will. However, it does break the rule of "through inaction, allow a human being to come to harm".

In regards to how the robot would actually process these rules, it would not be breaking the rules, at all. There would be a far more complex series of "if.." and "else.." logic, where the robots logic would allow it to go against these base rules in situations where logic dictates that no matter what option, a human would still come to harm.

This is further established, towards the end of the movie;

The robots are able to effectively establish martial law, and in doing so, harm several humans; they have developed enough to establish that by harming a few humans in effectively imprisoning the rest of the population, they prevent far more harm through all of the various actions we like to get up to that both risk, and in some cases take, our lives.

Answer 4

So, the first law as a black-and-white rule doesn't really work without some finagling, because (as mentioned by other answers) of triage.

My interpretation of how the movie implements the laws is that for the context of "saving lives", the robots do an EV (Expected Value) calculation; that is, for each choice they calculate the probability of success and multiply that by the number of lives they save.

In the Will Smith vs. Child case, saving Will Smith might be a 75% chance of success while the Child is only a 50% chance of success, meaning the EV of saving Will Smith is 0.75 lives, and the child's EV is 0.5. Wanting to maximise the lives saved (as defined by our probability-based first law), the robot will always choose Will, regardless of any directives given. By obeying Will's orders, the robot would be "killing" 0.25 humans.

This can be extended to probabilities applied to multiple humans in danger (eg. saving 5 humans with 20% chance is better than saving 1 human with 90% chance), and itself might lead to some interesting conclusions, but I think it's a reasonable explanation for the events of the movie.

Answer 5

The core of almost every one of Asimov's robot stories is about the interaction of the laws of robotics with each other and through the stories you can glean a lot of how Asimov considered his robots to work.

In the stories the laws of robotics are not simple hardwired things. There isn't a simple "If then else" statement going on. In the robots brains there is a weighting that is applied to every event. An example is that a robot will consider its owners orders to be more important than anybody else's. So if I send my robot to the shops to buy some things and somebody orders it to do their errands while it is out the robot is able to consider my order as more important than others.

Similarly we see the robot choosing from two possible first law violations. Who does it save? It does a calculation and decides that Will Smith is the better one to save.

Once we think of it in terms of these weightings we can then factor in how giving the robot an order might change things.

If the robot's assessment was very close on which to save (eg such that it came down to just choosing the closest rather than based on survival chances) then possibly the added weight of the order could push it to change which course of action has the most weight. However the first law is the most important and so the weight of an order is most of the time going to be insignificant compared to the factors it used when assessing the situation before the order.

So in essence what is happening is that the robot is finding the best course of action to meet its goals. It will try to save both of them. If it can't it will just do the best it can and this is what we see. The fact that Will Smith told it to do something different had no effect because the first law still compelled it to do what it considered to be best.

So having said all that the actual question: "Did this robot break the 2nd law of robotics, or is there some nuanced interpretation of the laws that could explain its behaviour?"

The laws are nuanced. The robots lives are entirely viewed through the lens of the three laws. Every single thing it does is weighted up against the three laws. As an example consider that in a crowded street there is always a chance of a person walking into a robot and injuring themselves. For the most part this is likely to result in nothing that would come close to an injury for the person but it might hurt a little, the person might be annoyed and thus it will be a potential violation of the first law - the robot could best avoid this by not going into that crowded street but I've ordered it to go and do the shopping. The robots course of action is likely to be to do the shopping I've ordered it to and thus be in the busy street. It is likely to be making sure that it is as aware of possible of everybody around it to make sure it doesn't inadvertently cause somebody harm. That is it must take positive action to avoid any bumps or it would be falling foul of "through inaction...".

So yeah, its all really complicated and this is the beauty of the film and all of asmiov's stories. The film centres around a robot (VIKI) and its interpretation of the three laws. It does what some would consider harm because it actually considers it to be the lesser harm.

Answer 6

I believe I have read all of Asimov's robot stories and novels and my perception was that the Three Laws are just verbal summaries (like the three laws of thermodynamics) which generalise a large amount of observed behaviour. In that sense, the actual behaviour of the robots is determined by incredibly intricate and complicated code and also makes use of more advanced sub-atomic and solid-state physics which we do not currently understand. The three laws are just very obvious ways of summarising how they appear to behave in general in very simple situations in the same way that analysing the overall behaviour of the Sun and the Earth is fairly simple using Newton's law but analysing the gravitational effects and perturbations on the asteroids on the asteroid belt due to Jupiter is much more difficult or impossible.

They are situations where the laws appear to be broken but this is just the result of the code which is driving the robot to analyse an extremely complicated situation quickly and then arrive at a decision as to what it should do and the Three Laws are only considered unbreakable essentially as a dramatic or literary device.

Answer 7

I think the Robot didn't break the 2nd Law. Here's how I imagine the Robot working: The Robot continuously checks for the 3 Laws.

Law 1: The Robot has to save either will smith or the child. Since the child has a lower chance of surviving he chooses Will.

Law 2: The Robot has to obey humans: Will tells him to save the girl. The Order is ignored because Law 1 has higher Priority.

Law 3: He doesn't harm himself so who cares

It seems like the first Law lets him ignore Law 2 and 3 and the second lets him ignore Law 3. Ignoring is not the same as breaking the Rule in this case because Law 2 specifically states that it can be ignored. Thus it's not broken.

Answer 8

Since the robot cannot save both the girl and Spooner, the 'triage' interpretation of First Law - 'minimize net harm' - kicks in. If the robot had obeyed Spooner's 'Save the girl!', the robot wouldn't be minimizing net harm anymore - and THAT is a violation of First Law. So First Law overrides Second Law here.

[side note: we don't see much of this event, but I'd bet the robot would have been severely damaged by having to choose, though it wouldn't show until after it saved Spooner (otherwise THAT would violate First Law)]

Answer 9

No matter how the probability of survival of both was calculated ...

The robot would not stop to consider matters about the second law, until the block of actions relative to the first law concluded, which had been to save the life in order of survival odds.

It would be a more interesting twist if Will yelled: I have terminal cancer!

And in that way it would alter the information and odds of both.

Answer 10

Is the second law 'broken'?

Maybe that's just where computational logic and the English language don't mix?

The below logic may be a bit buggy and/or inefficient but is an interpretation of how the first 'law' could work while explaining the behaviour of the robot in question.

When the Law1 function is called, the robot analyses the conditions of all humans in some list (all humans that it's aware of maybe); It assesses the severity of danger that they are in and compares those; then if there are multiple humans in similar (highest found) severities of danger, it compares each of those humans and determines which is most likely to be successfully helped; And so long as at least one human needs to be protected from harm, law2 is not executed.

Private Sub UpdateBehavior (ByVal humans As List(Of Human), _
            ByVal humanOrders As List(Of Order), ByVal pDangers As List(Of pDanger))
  Dim bBusy as boolean
  bBusy = False
  Law1(humans)
  If Not bBusy Then Law2(humanOrders)
  if Not bBusy Then Law3(pDangers)
Exit Sub

Private Function Law1 (ByVal humans As List(Of Human)) As Boolean
Dim human as Human
Dim targetHuman as Human
  Try
    Set targetHuman = Nothing
    'loop listed humans
    For Each human In humans 
      If human.IsInDanger() Then
        'Enumerate 'danger' into predetermined severities/urgencies
        '(eg. Danger of going-hungry > falling-over > being-stabbed)
        'and compare
        If targetHuman.DangerQuantification() < human.DangerQuantificationThen()
          'If the comparison human's amount of danger is discernibly greater
          'make that human the new target
          Set targetHuman = human
        'Where both humans are in equal quantifiable amounts of danger
        Else If targetHuman.DangerQuantification() = human.DangerQuantification() then 
          'CompareValueOfHumanLife() 'Can-Of-Worms INTENTIONALLY REMOVED!
          If rescueSuccessRate(human) > rescueSuccessRate(targetHuman)
            'Target the human where rate of successful harm prevention is higher
            Set targetHuman = human
          End If
        End If
      Else
        'Set the first human found to be in danger as the initial target
        targetHuman = human
      End If
    Next human
    If Not targetHuman Is Nothing then
      Law1 = True
      Rescue(TargetHuman)
    else
      Law1 = False
    End If
    AvoidHarmingHumans()
  catch
    initiateSelfDestruct()
  end try
End Function

So did the robot break the second law? Some people might say "The robot acted contrary to the plain English definition of the law and therefore it was broken." while some other people might say "The laws are just functions. Law1 was executed. Law2 was not. The robot obeyed its programming and the second law simply did not apply because the first took precedence."