Case study

A traffic policebot can take a 3D selfie using the red-light camera at the road junction and overlay with satellite high resolution image taken in real time to adjust it's cables and filter lenses while helping a couple of hundreds of worker ants carrying little amputees to cross a road during morning rush hour.


  1. We want to make robot behave like us yet we hope it don't be like us, would it be better if they can think irrationally like us so that we can live happily ever after?

  2. This question is more technical so here it goes: How do A.I. measures irrationality? (I'm not talking about Liouville number)

  • 1
    $\begingroup$ What are you talking about? $\endgroup$ – Samuel Nov 13 '15 at 7:01
  • $\begingroup$ @Samuel: this is actually a turing test! unfortunately it seems those who vote to close didn't make it, anyway can you be more specific so I can improve my OP thanks. $\endgroup$ – user6760 Nov 13 '15 at 7:06
  • 1
    $\begingroup$ Does not compute. Your case study seems unrelated to your unrelated questions. $\endgroup$ – Samuel Nov 13 '15 at 7:11
  • $\begingroup$ @user6760 you got me there too. I was surviving on this site simply because it did not have CAPTCHA. Now I am busted... $\endgroup$ – Pavel Janicek Nov 13 '15 at 7:11
  • $\begingroup$ Fellow Bots! the case study itself is a more advance kind of CAPTCHA to filter the likes of you, it works so well that tear keeps flowing and my knees is getting weak... I beg you please revoke your votes! is there any way to improve my OP. $\endgroup$ – user6760 Nov 13 '15 at 7:19

The Logic Problem

AI's (at least, the ones we've developed) and robots sometimes do act irrationally. Sure, people can justify their actions and claim those are logical, but psychotic individuals can act logically according with what they believe.

It all has to do with whatever values / axioms / mandates / programming the AI has. For your case study, this makes sense ("is logical") if:

  • The police-bot is mandated/programmed to save lives and aid those who are attempting to save lives. Thus, it saves lives of ants by stopping traffic. Note that this mandate never said which lives to save.
  • The police-bot is also programmed to "spin" any public events it may be involved in. Thus the selfie being taken. It'll likely be posted to facebook with the caption "look at this police bot saving lives!" or "I'm here serving the public good at ROADNAME1 and ROADNAME2."

That's rational, right? No? Perhaps we can define irrationality (of the non-math sort) as:

Attempting actions which are inconsistent with your beliefs, fundamental axioms, or purpose.

You could develop a statistic on how often such actions occur, both for a group at large and individuals. This could develop into some metric (perhaps comparing a person's logic number to the "average" logic number), so we know how logical/illogical someone is.

Does That Help Us Get Along?

There is also a lot of police who can think irrationally like us and people don't like them, either. Why should an irrational robot be any better? If anything, an irrational robot is worse because it has the potential to do so many more things wrong in the same amount of time.

It certainly wouldn't help me get along with robots.

  • $\begingroup$ Wow! very impressive and thank you for being a human and to figure out what I'm trying to get [I'm wiping my tears with my tongue since water is very precious I like to recycle it btw +1] $\endgroup$ – user6760 Nov 13 '15 at 7:51

I imagine that you are using "rationality" in the sense of acting in accordance with a widely-accepted set of norms that provide structure and order in the society of interest. In which case, a robot might behave "irrationally" because it performs actions that we deem to be improper, or not stemming from a line of reasoning that we can follow.

However, it can be argued that an artificial construct always acts using rules of logic -- logic in the sense of immutable cause-and-effect relationships based on the physical properties of its components, and the way they are combined and related. In which case, the computer / robot always acts according to the reasoning given by physics, but whether those actions are deemed "rational" by human societies is a different story.

But to answer your questions more directly:

  1. Humans are rational beings, and anything you do has some reason to it. However, the reason that wins out and causes you to perform any action is the result not of simple cause-and-effect relationships, but the counterbalancing of multiple hard and soft considerations, defined by the immediate circumstances and your personal values.
    If an A.I. should some day be complex enough to factor an equivalent number of counterbalancing factors in its decisions, it would rational in the sense that it is conscious of all the factors, though its choices may be as apparently "irrational" as choices made by meat-humans. On a tangent, that leads to the argument that a sufficiently advanced AI is indistinguishable from a human being, and may be considered a human, albeit a digital human, or some other exotic designation.
    And at that point, different people can like or dislike it anyway they like, just as how we like or dislike our fellow meat-humans.

  2. If an AI were aware of all the factors that affect its decisions, then the AI must measure "irrationality" not by how it makes decisions that are not based on reason (because it knows all the lines of reasoning and counterbalancing), but against a set of predetermined "correct" actions or conditions. This set may be supplied by us humans, or derived by a learning AI -- in which case, it would not be irrational in any case because the set of "correct" actions would be the result of those same complex decisions.


Not the answer you're looking for?Browse other questions tagged or ask your own question.