Automated argument adjudication to solve ethical problems in multi-agent environments

Abstract

Abstract Suppose an artificial agent <m:math xmlns:m="http://www.w3.org/1998/Math/MathML"><m:msub><m:mrow><m:mi>a</m:mi></m:mrow><m:mrow><m:mtext>adj</m:mtext></m:mrow></m:msub></m:math> {a}_{\text{adj}} , as time unfolds, (i) receives from multiple artificial agents (which may, in turn, themselves have received from yet other such agents…) propositional content, and (ii) must solve an ethical problem on the basis of what it has received. How should <m:math xmlns:m="http://www.w3.org/1998/Math/MathML"><m:msub><m:mrow><m:mi>a</m:mi></m:mrow><m:mrow><m:mtext>adj</m:mtext></m:mrow></m:msub></m:math> {a}_{\text{adj}} adjudicate what it has received in order to produce such a solution? We consider an environment infused with logicist artificial agents <m:math xmlns:m="http://www.w3.org/1998/Math/MathML"><m:msub><m:mrow><m:mi>a</m:mi></m:mrow><m:mrow><m:mn>1</m:mn></m:mrow></m:msub><m:mo>,</m:mo><m:msub><m:mrow><m:mi>a</m:mi></m:mrow><m:mrow><m:mn>2</m:mn></m:mrow></m:msub><m:mo>,</m:mo><m:mi>…</m:mi><m:mo>,</m:mo><m:msub><m:mrow><m:mi>a</m:mi></m:mrow><m:mrow><m:mi>n</m:mi></m:mrow></m:msub></m:math> {a}_{1},{a}_{2},\ldots ,{a}_{n} that sense and report their findings to “adjudicator” agents who must solve ethical problems. (Many if not most of these agents may be robots.) In such an environment, inconsistency is a virtual guarantee: <m:math xmlns:m="http://www.w3.org/1998/Math/MathML"><m:msub><m:mrow><m:mi>a</m:mi></m:mrow><m:mrow><m:mtext>adj</m:mtext></m:mrow></m:msub></m:math> {a}_{\text{adj}} may, for instance, receive a report from <m:math xmlns:m="http://www.w3.org/1998/Math/MathML"><m:msub><m:mrow><m:mi>a</m:mi></m:mrow><m:mrow><m:mn>1</m:mn></m:mrow></m:msub></m:math> {a}_{1} that proposition <m:math xmlns:m="http://www.w3.org/1998/Math/MathML"><m:mi>ϕ</m:mi></m:math> \phi holds, then from <m:math xmlns:m="http://www.w3.org/1998/Math/MathML"><m:msub><m:mrow><m:mi>a</m:mi></m:mrow><m:mrow><m:mn>2</m:mn></m:mrow></m:msub></m:math> {a}_{2} that <m:math xmlns:m="http://www.w3.org/1998/Math/MathML"><m:mo>¬</m:mo><m:mi>ϕ</m:mi></m:math> \neg \phi holds, and then from <m:math xmlns:m="http://www.w3.org/1998/Math/MathML"><m:msub><m:mrow><m:mi>a</m:mi></m:mrow><m:mrow><m:mn>3</m:mn></m:mrow></m:msub></m:math> {a}_{3} that neither <m:math xmlns:m="http://www.w3.org/1998/Math/MathML"><m:mi>ϕ</m:mi></m:math> \phi nor <m:math xmlns:m="http://www.w3.org/1998/Math/MathML"><m:mo>¬</m:mo><m:mi>ϕ</m:mi></m:math> \neg \phi should be believed, but rather <m:math xmlns:m="http://www.w3.org/1998/Math/MathML"><m:mi>ψ</m:mi></m:math> \psi instead, at some level of likelihood. We further assume that agents receiving such incompatible reports will nonetheless sometimes simply need, before long, to make decisions on the basis of these reports, in order to try to solve ethical problems. We provide a solution to such a quandary: AI capable of adjudicating competing reports from subsidiary agents through time, and delivering to humans a rational, ethically correct (relative to underlying ethical principles) recommendation based upon such adjudication. To illuminate our solution, we anchor it to a particular scenario.