[A presentation on this today from CybSoc’s 50th anniversary seemed strangely naive in some ways, but generated lots of discussion and thinking]

The Ethical Regulator Theorem

The ethical regulator theorem provides a basis for systematically evaluating the adequacy of existing or proposed designs for systems that make decisions that can have ethical consequences; regardless of whether the regulating agents are humans, machines, cyberanthropic hybrids, organizations, corporations, or government institutions.

The theorem builds upon the law of requisite variety and the good regulator theorem to define nine requisites that are necessary and sufficient for a cybernetic regulator to be both effective and ethical:

Purpose expressed as unambiguously prioritized goals.
Truth about the past and present.
Variety of possible actions.
Predictability of the future effects of actions.
Intelligence to choose the best actions.
Influence on the system being regulated.
Ethics expressed as unambiguously prioritized rules.
Integrity of all subsystems.
Transparency of ethical behaviour.

Of these nine requisites, only the first six are necessary for a regulator to be effective. If a system does not need to be ethical, the three requisites ethics, integrity, and transparency are optional.

Super-Ethical Systems

A six-level framework is proposed for classifying cybernetic and superintelligent systems, which highlights a future time-line bifurcation that results in one of two mutually exclusive outcomes:

The human race is protected by superintelligent, ethically adequate “Super-Ethical” systems.
The human race is dominated by superintelligent, ethically inadequate “Super-Unethical” systems.

For more information, see the PDF: Cybernetics 3.0: Ethical Systems.

W. Ross Ashby Digital Archive Mirror

This domain also hosts a mirror copy of the W. Ross Ashby Digital Archive for use when the original archive site is unavailable because it has been permanently screwed-up with monetizing malware by Site5, the crap hosting company, and then Error 500’ed to death 😦

Source: Ethical Regulator Theorem

mandospags 8:33 am on September 23, 2018

The best thing about the presentation is that it was a good session to have after lunch and certainly raised the energy level in the room!

I think that it would be a complete disaster if the government of the day could decide on (and remotely update) the ethical rules that a computer would follow. (Let’s ignore for a moment the weakness of the suggestion that a rule based AI can be an AI, or that following rules is all that it takes to be ethical)
abbeboulah 8:57 am on September 28, 2018

This system of ethics seems to have some blind spots — some even regarding the most important purposes of ethics — the interaction with other beings and systems. Some questions, thoughts and aspects that should be added to this picture. Questions that highlight the urgency of raising the issue: Thanks!
For starters:
1: Purpose:
a) Don’t participant agents even within the system, but obviously those ‘outside’, in other systems, have different pursuits? So the question of how to deal with those differences is a key issue — which could even be expressed as the question of predictability — how can we realistically expect others to behave, and what behaviors of mine (‘our system’ ) should those others be able to count on in our interactions?
b) For those aspects of our lives that can be described as ‘exploration’ or discovery: can the purpose be so clearly and unambiguously defined? Will every new discovery not change our purposes?

2: Truth
a) Truth is a noble goal one would not wish to argue with. — But even rigorously established and corroborated scientific hypotheses are considered more or less probable, not 100% ‘true’ open to improvement or new paradigms even for ‘laws of nature’, and even more so in social systems? The ‘facts’ of past events are hard enough to pin down with reliable accuracy — but the causes and influencing factors that brought them about remain open to endless speculation and ‘interpretation’: stories, myths, allegories whose very presentation as ‘not factually true’ but expressing (selective and purposefully exhortative) deeper truth?
b) The item 4 ‘Predictability’ further down the list is a valid and important qualification of the truth principle, but does not prevent many from insisting that the ‘truth’ (or Big Data) of past and present will be enough to determine the recommended future plans. The missing aspect for pour plans for the future is not truth but the desirability, validity, goodness of the goal we claim to pursue with some action. I usually summarize all those in the term ‘plausibility’. Here, even some, the plausibility of conflicting goals of others are the hurdles of ethics.

3: Variety of possible actions.
Yes. But aren’t’ there always at least two options: ‘do something’ or ‘do nothing’?

4: Predictability of future effects.
This seems to imply that these things are indeed predictable and therefore should be predicted, in order for actions to count as ethical? Would it, in may cases, not be more honest (and thus more ethical) to acknowledge that our ability to predict is woefully inadequate — that we are going to haven to gamble, take chances? That even the most carefully thought out and simulated consequences of elaborate plans will produce efforts — by agents who want to ‘make a difference’ — to precisely change those predicted effects?

5: Intelligence.
Would be nice. Choosing the ‘best’ actions is a noble goal, the task of pinning down what’s ‘best’ is the sticking point. But are we not expecting even agents with less intelligence (ability to absorb pertinent information and convert it into effective action, as one crude definition) to behave ‘ethically’ — according to some principle that by definition (‘primary, first consideration’) overrides any other intelligence speculations and optimization efforts? Do the right thing? ‘My country, right or wrong’ — maybe questionable from an ‘intelligence’ point of view, but widely held to be a quite ‘ethical’ stance?

6: Influence
Implication If I don’t have (or believe I have) any influence on a given system, I can claim to be ethical?

7: Ethics expressed as unambiguously prioritized rules
Is there a question about circularity definition here? The none requisites for ethical regulation requiring ethics. And some of the questions about the other requisites putting a big question mark on the unambiguosity and prioritizing aspects?

8 Integrity of all subsystems. This requires some more thought and explanation. If it’s a ‘necessary’ condition, would somebody claiming a a lack of ‘integrity’ in some subsystem destroy the entire system’s ethics?

9 Transparency.
In any social interaction that can be adequately described as a kind of ‘game’ — isn’t the lack of transparency of players intentions making the very interest of the game (to figure out the ‘opponent’s’ hidden strategy and develop contingency tactics to thwart it, that must be equally kept hidden for the opponent)? Sure, in a ‘friendly game’, according to the rules of the game. But without that challenge to discover the hidden strategies, the game would be uninteresting and boring. Now, for ‘games’ against wicked global challenges with both natural and human agents, would it not be violating item 5 in the list to disregard the possible need for a differentiated stance regarding this requisite?
- abbeboulah 1:08 pm on September 28, 2018
  
  Corrrection: “Implication If I don’t have (or believe I have) any influence on a given system, I can claim to be ethical?” Should be ‘CANNOT claim…?