The Logic of Ethics

How difficult will it be to translate ethics into logic trees that can be encoded into software for use by AIs? In a simplified ethical decision, there is a starting state, usually at least two possible actions (or an action/inaction pair) to choose between and outcome states to calculate and compare. In order to make an ethical decision, the action that produces the most desirable outcome state must be selected. We’ll need to create representations of the starting state, the actions to consider, the possible outcome states and the decision branching.

Representation of state:
In order to create a representation of “ethical state” for both a starting place and and a theoretical end state, a collection of attributes will need to be chosen that can accurately reflect an ethical state. These should include (among others): a scope factor (greatest number), a depth factor (greatest good) and any “boundary conditions” such as moral codes, laws and other regulatory issues. A standard coefficient scale needs to be established in order to allow comparison. A summation algorithm or process will be needed to create an overall measurement of “ethical state” once all the attributes are filled in.

Decision description:
The action decision choices change the starting state and we will need to describe those changes in terms of the attributes we’ve decided to use in describing the state above. In a very simple theoretical case, this might offer an instant suggestion for the best ethical decision because as the summations are performed, the results can be compared. In a more complex case, it may be difficult to calculate ripple effects across the scope and depth matrix.

Decision branching logic:
While a simple case might not need any branching; more complex, real life scenarios may need some re-iterative application, specially when searching for an optimum decision description. A key design consideration will be whether to use all binary logic (yes/no) or multi-valued decision outcomes. Multi-valued or analog logic is more complex but offers better modeling capabilities. Evolutionary computational techniques such as genetic algorithms and genetic programming are likely to be useful.


SEE ALSO:
Ethics vs Morals
Evolutionary Computation

Leave a Reply

You must be logged in to post a comment.