Showing posts with label Alan. Show all posts
Showing posts with label Alan. Show all posts

Artificial Intelligence - What Is The The Moral Turing Test, Or Ethical Turing Test?


The Moral Turing Test, also known as the Ethical Turing Test. 

Ethical (Moral) Turing Test, or MTT, is a variant of the Turing Test created by Alan Turing (1912–1954), a mathematician and computer scientist.

A human judge uses a series of written questions and replies to try to tell the difference between a computer program and a person.

If the computer program imitates a person to the point that the human judge cannot discern the difference between the computer program's and the human's replies, the program has passed the test, indicating that it is capable of intelligent reasoning.

The Moral Turing Test is a more precise version of the Turing Test that is used to assess a machine's ethical decision-making.

The machine is initially taught broad ethical standards and how to obey them.

When faced with an ethical problem, the computer should be able to make judgments based on those ethical standards.

The choices of the computer are then contrasted to those of a human control, usually an ethicist.

The Moral Turing Test is usually only used in certain settings that are relevant to a specific area of study.

If the machine is presented with an ethical problem about health care, for example, its choice will be compared to that of a human health-care professional rather than a generic human control.

The Moral Turing Test has been regarded as a flawed method of determining a machine's capacity to exercise moral agency.

The Turing Test uses imitation to determine if a computer can think, but detractors of the Moral Turing Test argue that in an ethical issue, imitation may be performed by misleading replies rather than moral thinking.

However, some say that morality cannot be determined only on the basis of vocal reactions.

Rather, in a classic Turing Test, the judge must be able to see what is going on in the background—the reasoning, analysis of alternatives, decision-making, and actual action—all of which would be concealed from view.

The comparative Moral Turing Test (cMTT), the Total Turing Test, and verification are all alternatives and modifications to the Moral Turing Test.

In a comparative Moral Turing Test, the judge compares the machine's narrated acts to a human control rather than its spoken replies.

In a Total Turing Test, the judge may see the machine's real activities and interactions in comparison to the human control.

Verification takes a different approach than testing, concentrating on the method behind the machine's reaction rather than the result.

Verification is assessing the design and performance of the machine to determine how it makes decisions.

Verification proponents argue that focusing on the process rather than the outcome acknowledges that moral questions rarely have a single correct answer, and that the process by which the machine arrived at an outcome reveals more about the machine's ability to make ethical decisions than the decision itself.

~ Jai Krishna Ponnappan

Find Jai on Twitter | LinkedIn | Instagram

You may also want to read more about Artificial Intelligence here.

See also: 

Turing, Alan; Turing Test

References & Further Reading:

Arnold, Thomas, and Matthias Scheutz. 2016. “Against the Moral Turing Test: Accountable Design and the Moral Reasoning of Autonomous Systems.” Ethics and Infor￾mation Technology 18:103–15.

Gerdes, Anne, and Peter Øhrstrøm. 2015. “Issues in Robot Ethics Seen through the Lens of a Moral Turing Test.” Journal of Information, Communication, and Ethics in Society 13, no. 2: 98–109.

Luxton, David D., Susan Leigh Anderson, and Michael Anderson. 2016. “Ethical Issues and Artificial Intelligence Technologies in Behavioral and Mental Health Care.” In Artificial Intelligence in Behavioral and Mental Health Care, edited by David D. Luxton, 255–76. Amsterdam: Elsevier Academic Press.

What Is Artificial General Intelligence?

Artificial General Intelligence (AGI) is defined as the software representation of generalized human cognitive capacities that enables the ...