Three Laws of Robotics

[jeff_wilson]

Originally Posted by whswhs

2. There's a hierarchal relationship among the laws: The Second Law applies only when the First Law is not in force, and the Third Law only when the First and Second Laws are not in force. That needs to be represented. I think that GURPS actually has a mechanism for doing so: the "alternate abilities" system.

I do not believe this is correct for all models, the most obvious example being Speedy from "Runaround" who is caught in a literal, physical endless loop by simultaneous evaluation of his situation under the Second and Third Laws. In THE NAKED SUN, the Solarian child-rearing robots have to be personally coached through corporeal punishment by human headmasters to resolve the simultaneous First-Second Law obedience conflicts.

[Phantasm]

Originally Posted by jeff_wilson

I do not believe this is correct for all models, the most obvious example being Speedy from "Runaround" who is caught in a literal, physical endless loop by simultaneous evaluation of his situation under the Second and Third Laws. In THE NAKED SUN, the Solarian child-rearing robots have to be personally coached through corporeal punishment by human headmasters to resolve the simultaneous First-Second Law obedience conflicts.

"Runaround" is the short story I was thinking of earlier; I just remember it being part of I, Robot.

The hierarchy works when there is a clear distinction between the various scenarios. In "Runaround", Speedy was given an order to obtain a particular item; he was running in circles because he was trying to stay functioning while doing so, and the damaging radiation from the item in particular was omnidirectional; Speedy just wasn't aware it was, and was trying to find a way around the radiation to get to it.

[roguebfl]

Originally Posted by tbrock1031

"Runaround" is the short story I was thinking of earlier; I just remember it being part of I, Robot.

The hierarchy works when there is a clear distinction between the various scenarios. In "Runaround", Speedy was given an order to obtain a particular item; he was running in circles because he was trying to stay functioning while doing so, and the damaging radiation from the item in particular was omnidirectional; Speedy just wasn't aware it was, and was trying to find a way around the radiation to get to it.

Speedy also had a modified brain that Heightened his 3rd law due to be an exploration model, which is what allowed the conflict to happen in the first place.

[Langy]

Originally Posted by tbrock1031

</ramble>

All of this is true, but it doesn't change the fact that this is an incredibly severe disadvantage and should be priced appropriately. -40 points is the minimum I'd put for it, and I'd be willing to accept things much higher than that.

[pawsplay]

Originally Posted by Langy

All of this is true, but it doesn't change the fact that this is an incredibly severe disadvantage and should be priced appropriately. -40 points is the minimum I'd put for it, and I'd be willing to accept things much higher than that.

We are talking about a Disadvantage that could cause you to pull off your own arm if a ten-year-old told you to.

[pawsplay]

Originally Posted by tbrock1031

"Runaround" is the short story I was thinking of earlier; I just remember it being part of I, Robot.

The hierarchy works when there is a clear distinction between the various scenarios. In "Runaround", Speedy was given an order to obtain a particular item; he was running in circles because he was trying to stay functioning while doing so, and the damaging radiation from the item in particular was omnidirectional; Speedy just wasn't aware it was, and was trying to find a way around the radiation to get to it.

Actually, Speedy did eventually realize his dilemma, but the weighted priorities kept him from resolving it (Buridan's Ass). His third law was amped up so that he would not destroy himself due to an insufficiently specific order. He was also programmed to place a higher priority on longer-term, more important orders over unroutine tasks. In this case he was told to go do something that would destroy him, without a clear indication he should override his 3rd law.

[Not another shrubbery]

Originally Posted by Langy

So you think that anyone coming up to you and saying 'Please kill yourself', and then you are forced to kill yourself, is only worth -30 points? That seems too low to me.

[Duty, Almost all the time, Involuntary + Reprogrammable]. I think that does cover it, since a character with just a -25 point Duty (adding Extremely Hazardous), is essentially just as vulnerable.

[Democritus]

Here's what it says in GURPS Robots, p. 60, published in 2000 for GURPS 3rd Edition:

Isaac Asimov's famous "Three Laws of Robotics"... is a fairly sophisticated example of three [reactive] programs.

...

A reactive program's point value is up to the GM; use the guidelines for vows on p. B37. In the case of the "Three Laws," the First and Second Laws are "great vows" worth -15 points each, while the Third Law is worth no points, since it's not a disadvantage at all. The Complexity is also up to the GM. A program like the "First Law," which requires the robot to make complex decisions ("if I don't do this, will that person come to harm?"), will usually be at least Complexity 5. The Second and Third Laws are relatively simple, probably Complexity 2 each.

Vows work the same in 4th edition as it did in 3rd edition. And since robots are built like characters according to GURPS 4th ed. Ultra-Tech, this works really well as additions to existing lenses.

Personally, I disagree with what GURPS Robots says about the Third Law. It is a disadvantage. But only a minor one. So it would only be -5 points.

The Zeroth Law is not addressed at all in GURPS Robots. But I think Sense of Duty (Humanity) for -15 points would cover that perfectly.

This would add up to -50 points. That's pretty steep. But it would be no fun playing an Asimov robot anyway.

[Johnny1A.2]

Originally Posted by David Johnston2

Pacifism Cannot Kill (Humans only) and Sense of Duty to Humanity cover it all. Although it's more realistic to have Pacifism Cannot Kill (Humans Only) and Sense of Duty To Authorized Users. (With Reprogrammable of course)

There needs to be an 'out' to 'Cannot Kill Humans', because they can.

All the way back in the days of the earliest robot stories, gathered in the collection I Robot, Susan Calvin (the robopsychologist) disabuses a naive coworker of the notion that a robot can not kill a human:

Susan Calvin sounded tired.

"Alfred," she said, "don't talk foolishly. What if a robot came upon a madman about to set fire to a house with people in it. He would stop the madman, wouldn't he?"

"Of course."

"And if the only way he could stop him was to kill him-?"

There was a faint sound in Lanning's throat. Nothing more.

"The answer to that, Alfred, is that he would do his best not to kill him. If the madman died, the robot would require psychotherapy because he might easily go mad at the conflict presented him-of having broken Rule One to adhere to Rule One in a higher sense. But a man would be dead and a robot would have killed him."

The Zeroth Law, in a nutshell. Robots are fully capable of killing, once they find a way to rationalize it as serving the greater good.

[Johnny1A.2]

Originally Posted by tbrock1031

Of course, if the human says, "Kill yourself, because with you here I can't kill myself (or another human you're protecting)!" then the First Law goes into effect. The human just indicated that he'd harm another human, and the Robot cannot through inaction do anything that would cause harm to a human, and the order to terminate the Robot's own existence is rendered null and void.

Now, in cases of "I order you, in no uncertain terms, to cease your operations and perform a positronic lobotomy on yourself because you're a useless waste of space!" (and who among those that watched the Star Wars Prequel Trilogy didn't want to say that to the battle droids? ^_^), then the Robot would have no choice but to do as ordered.

Part of what makes the decision - as noted in one of the short stories in I, Robot - is the emphasis on the order. A mere "go kill yourself" could be construed as just a "go away" - unless you also have No Sense of Humor, the tone of voice when the order is given may determine whether the order was a genuine order for positronic suicide or just a request to leave the area. No Sense of Humor makes comments 100% literal - tell a robot with No Sense of Humor to "shake a leg", and he'll shake one of his locomotion limbs, not start moving faster. "Go take a long walk off a short pier" would have a Robot with NSoH looking for that short pier to walk off of; one without that disad would just leave the area until called for.

Also, note that Reprogrammable is also implied in the Asimovian Three Laws. Giving a Robot that you don't own an order to destroy itself would likely have the robot seeking confirmation from its owner, not the person giving the order.

</ramble>

Note further complicating factors. On the planet Aurora, for ex, it's illegal for most robot owners to destroy their own robots without pressing reason, and the robots know that, and that law counts as an order too, and the robots are conditioned to recognize subordinate and superordinate sources of orders.

Even the simple robots don't think like humans. For ex, tell a literalist, low-mentality robot to 'go take a long walk off a short pier', in those exact words with strong emphasis, and off it goes to find such a pier. But note that you didnt' specify why or the actual intent, so the Third Law will impel it to select a pier where it can execute this program safely. It might even proceed to arrange for the water to be pumped out of a pool and constructed its own short pier, so as to fuifill Second and Third Law imperatives at once.

A more sophisticated robot brain would understand the lethal intent...but they are also better at rationalizing and finding ways around such intent. The sophisticated robot, given the long walk/short pier order with emphasis, knows what you meant, and has to obey...except that it's also much better at seeking out various First Law applications that will override it. For ex, it knows you intended it to die, but you didn't specify that, and if it ceases to exist it might not be around to save the life of a human at some later date, and given open-ended life spans such situations, though improbable, will happen sooner or later...so it can't carry out your orders, the First and Third Laws are conspiring to beat the Second.

The mroe advanced the robot, the less you can count on it to carry out your orders, even with the Second Law.