(Immanuel Kant) How can computers solve ethical problems?

logos

If Kant views ethics as a logical problem,
and if machines are made to solve logical problems,
then should machines be able to solve ethical problems as logical problems?

Artificial intelligence has an ethics problem. There are many GANs that produce ethically terrible content, like deep fakes. But the internet depends on recommendation systems for advertising and news feeds. I'm trying to convert Kant's categorical_imperative into hard code as an attempt to solve this problem, so computers can tell the difference themselves. It's purely theoretical.

Consider the crudely written prototype below.

    function mass_adoption(action_input) { 
    	if unknown 
    		   return unknown
        else if precedent == True 
                if precedent == Ethical
                    return output_goal_variable
        else
                return "Not Ethical"
    }
    
    function categorical_imperative( goal_Variable_input, action_Variable_input) {     
        let goal_Variable = goal_Variable_input
        let action_Variable = action_Variable_input
        
        define output_goal_variable = function mass_adoption(action_Variable)

        if  output_goal_variable == goal_Variable
            then 
            	return True, "actions obtains goals through mass adoption"
            else
                return False, "actions through mass adoption prevents goal"
                    
        }

examples:
The Goal is to write a book using these available Actions: write based off true life, write based off Lord Of The Rings, or write based off Nordic folklore. Using the categorical_imperative function, what would the return be on each action and why?

    function categorical_imperative(write_a_book, write_based_off_truelife)
        returns true, "because writing a book will result in the mass adoption of more books being written from true life."

    function categorical_imperative(write_a_book, write_based_off_Lordoftherings)
        returns false, "because mass adoption of plagiarism will return in more lawsuits instead of books being written."  

    function categorical_imperative(write_a_book, write_based_off_Nordicfolklore)
        returns true, "because writing a book with stories from the public domain is legal, thus resulting in more books being written."

*edited for formatting*

Echarmion

If Kant views ethics as a logical problem,
and if machines are made to solve logical problems,
then should machines be able to solve ethical problems as logical problems? — logos

Well the first question is, does Kant view ethics as a "logical problem"?

The Goal is to write a book using these available Actions: write based off true life, write based off Lord Of The Rings, or write based off Nordic folklore. Using the categorical_imperative function, what would the return be on each action and why? — logos

I am not seeing anything about maxims here. The categorical imperative is not a tool to judge single actions, but to judge maxims.

alcontali

If Kant views ethics as a logical problem,
and if machines are made to solve logical problems,
then should machines be able to solve ethical problems as logical problems? — logos

A formal system of ethics will have system-wide premises, if only, the axiomatization of logic itself.

Unless you adopt an extreme form of logicism, the axiomatization of logic itself is insufficient to address ethical problems. Therefore, you will need to add the system-wide premises of your ethics in your formal system.

Stephen Wolfram believes that it is possible to discover theorems, i.e. "answer questions", in a formal system by mechanical enumeration. His (simple) example is boolean algebra. I personally do not believe that it will (generally) be possible. For example, it took 350+ years to discover the proof for Fermat's Last Theorem entailing from the system-wide premises of number theory. I do not believe that a machine would have discovered it.

However, I do believe that mechanical verification of proof within a formal system is attainable. Hence I do not believe that machines will "solve" ethical problems but I do believe that machines can "verify" that a proposed solution is indeed a legitimate solution.

So, in my opinion, machines should be able to verify the solution for ethical problems as logical problems.

tim wood

However, I do believe that mechanical verification of proof within a formal system is attainable. Hence I do not believe that machines will "solve" ethical problems but I do believe that machines can "verify" that a proposed solution is indeed a legitimate solution. — alcontali

A restatement of P v. NP. But I wonder if this is really an NP (or worse - probably much worse) problem. For example, ethical problems are often (usually? always?) contextual, even personal, problems. That is, right for the situation/right for the person. How can a computer handle that, being itself neither contextual nor a person?

alcontali

A restatement of P v. NP. But I wonder if this is really an NP (or worse - probably much worse) problem. For example, ethical problems are often (usually? always?) contextual, even personal, problems. That is, right for the situation/right for the person. How can a computer handle that, being itself neither contextual nor a person? — tim wood

If you look at a practical example of a formal system for morality, it seems to be possible to generally answer a good number of questions, e.g. https://islamqa.info in morality.

The answers are produced as syntactic entailments from scripture.

In my opinion, it should be possible to mechanically verify a syntactic entailment using a proof assistant such as Coq or Isabelle. The difficulty consists in encoding the scriptural base as well as the advisories in the tool's formal language.

Hence, this kind of project would consist in going through the existing database of advisories and re-encode them. That is a lot of work but I think it should be possible.

tim wood

↪alcontali

Sure, possible. Ethics by the numbers. Hmm. Scriptural. Which one? And haven't you heard, both Islam and Christianity are down on sexual deviance, including women's rights. Or maybe Nicomachean Ethics? Or Stoic? And that's just the West. And so on & etc.

I have no doubt that a data base of ethics that could inform and answer could be very useful - like IBM's Watson - another interesting part of the online experience. But as authoritative as to action to be taken? I'm not ready for that, are you?

Another consideration: ethics is essentially creative, a reaction to the now. A data base isn't. A analogy of sorts: Imagine it was decided that humanity had no need of any numbers not already identified. That's a lot of numbers. But you've posted to maths threads here; you know better than I that for all the numbers known, a whole lot more are unknown - would you be willing to give those up and rely on just those known?

Certainly many known numbers would suffice for many things, just as many sets of ethics could cover many situations. Do you think something like that is a good idea?

alcontali

Sure, possible. Ethics by the numbers. Hmm. Scriptural. Which one? And haven't you heard, both Islam and Christianity are down on sexual deviance, including women's rights. — tim wood

For the one or the other totally incomprehensible reason, there are western people who are totally deluded to believe that their take on women's rights would be universal. That is a complete delusion and even a dangerous one.

We totally disagree on that subject, and it is not even negotiable. In fact, even in the West there are entire popular movements, such as the red-pill philosophy who beg to disagree on the subject. Youtube is full of videos on why and how they disagree.

Another consideration: ethics is essentially creative, a reaction to the now. A data base isn't. A analogy of sorts: Imagine it was decided that humanity had no need of any numbers not already identified. That's a lot of numbers. — tim wood

The database of mathematics also keeps growing.

So how big is the historical corpus of mathematics? There’ve probably been about 3 million mathematical papers published altogether—or about 100 million pages, growing at a rate of about 2 million pages per year. And in all of these papers, perhaps 5 million distinct theorems have been formally stated. — Stephen Wolfram, Computational Knowledge and the Future of Pure Mathematics

Each theorem is a number and each proof is an associated number. So, yes, that is a lot of numbers, (theorem,proof) tuples actually. Many undiscovered numbers are undoubtedly uninteresting.

Interestingness. Of course, the general problem of ranking “what’s interesting” comes up all over Wolfram|Alpha. — Stephen Wolfram, Computational Knowledge and the Future of Pure Mathematics

Unfortunately, in his article, Stephen Wolfram does not really make progress in explaining how to distinguish between "interesting" and "uninteresting". He merely acknowledges the problem.

Certainly many known numbers would suffice for many things, just as many sets of ethics could cover many situations. Do you think something like that is a good idea? — tim wood

I see the solution as a growing databases similar to the collection of theorems on number theory or set theory. In the case of morality, it would be rulings within a database of Jewish law or one with rulings within Islamic law, stored with verifiable proofs from scripture. The problem is that they need to be encoded in the database's formal language. That is relatively hard and also a lot of work.

Stephen Wolfram advocates doing that for the existing corpus of mathematics:

What would unquestionably be worthwhile, however, is to put the theorems into a genuine computable form: to actually take theorems from papers and rewrite them in a precise symbolic language.

Will it be possible to do this automatically? Eventually I suspect large parts of it will. Today we can take small fragments of theorems from papers and use the linguistic understanding system built for Wolfram|Alpha to turn them into pieces of Wolfram Language code. But it should gradually be possible to extend this to larger fragments—and eventually get to the point where it takes, at most, modest human effort to convert a typical theorem to precise symbolic form. — Stephen Wolfram, Computational Knowledge and the Future of Pure Mathematics

I personally do not believe that it can be done automatically, not even just "large parts". He also exaggerates the ability of Wolfram|Alpha to encode natural language into formal language. In my opinion, it won't require "at most, modest human effort". I think that it will be an enormous amount of work to do that.

Still, I do agree that it will be worthwhile. The most important reason why nobody is currently working on a curated corpus of mathematics, is because the corporate academic-publishing oligarchy has hijacked many of the copyrights on mathematical publications.

(Immanuel Kant) How can computers solve ethical problems?

Welcome to The Philosophy Forum!

Categories

More Discussions