About Intelligent-Agent Ethics, Meta-Ethics and ... Computational Axiology: a  Socio-Cognitive Perspective 
on Meta-Knowledge Engineering Server.   ENEA, CAMO,HID  (High-Intelligence and Decision Research Group), 2001/3
Keywords: ethics, socio-cognitive science and engineering, intelligence, axiology, preferences (IPK paradigm), agents
MKS (Meta-Knowledge Server)    |  AGENTS and INTELLIGENCE  |    Meta-Ethics |   Computational Axiology
MEFIT: Meta-Ethics Frameworks of IT: Socio-Cognitive Approach PARTNER SEARCH

Ethics of/in Intelligent Agent Technology -  initial opinions
 Copy from:http://www.csee.umbc.edu/agentslist/archive/1996b/0431.html
From: gadomski adam (gadomski_a@casaccia.enea.it)

Fri, 7 Mar 1997 11:00:31 +0100
  • Messages sorted by: [ date ][ thread ][ subject ][ author ]
  • Next message: Ian J. Dickinson: "Re: Agents that Lie and Steal?"
  • Previous message: Michael Stillger: "Re: Agents that Lie and Steal?"
Dear Gerd

At 12.50 05/03/97 MET, Gerd Wagner wrote:
Dott. Adam Maria Gadomski wrote

The real problem is not technical one, it is localized on the level of personal ethics of the developers and end-users. Summarizing a "lying, cheating and stealing" - no problem for designers, and as always,  a problem for humanity, a problem of the power, money and interests.

 I disagree. Cheating and stealing may be ethical topics, but lying is only in some cases malicious behavior, in others it is dictated by rationality. I quote from a recent paper of mine:

Whenever agents deal with confidential information, it is  important that they comply with a principled security policy.
We show how the database concept of multi-level security can be applied to inter-agent communication.
This includes the case where an unauthorized agent is misinformed on purpose in order to protect confidential information. (from my(Gerd Wagner) paper "Multi-Level Security in Multiagent  Systems", ...

- Well, you omitted one part of my comment ( Gadomski,27/02/97) which is the following:

Of course, the problem is not trivial if the situation is complex BUT the complex problem solving in a complex situation is ALWAYS complex*complex.
Therefore, from  the knowledge engineers perspective, the real problem is not a type of mental activity we intend to implement but how to formalize  rational goal-oriented activities, in general.

- In this context, of course, you can have many practical problems but the key ethical problem is shifted on the higher meta-level.

- Does the decision (rule) "an unauthorized agent is misinformed on purpose in order to protect confidential information" be ethically correct (?)

- This is the ETHICAL problem of intelligent agent developers.

- For me, this rules can be justified in specific circumstances, but is never ethic... OK, this is a meta-ethic topic,... we can have different ethics.

 Lying in order to protect confidential information is *not malicious*  behavior, it's  *not even egoistic*, but ethically justified because the protection of sensitive information is a high value in its own and it may be necessary to save lives or protect fundamental rights of persons.

- Here , your motivation is dangerous because it is essentially incompleted, we, as users and causually "unauthorized agents", should be protected
against such reasoning, You should add - Who justifies that an particular "lying" can be the necessary and sufficient condition
"to save lives or protect fundamental rights of persons"?
As well as, who will be responsible of unexpected consequences of this false information?

Therefore we have a technical problem here: we have to define secure query answering, which includes the case of lying, in accordance with the rationality principle of minimal misinformation.

- I would like to know "the rationality principle of minimal misinformation". Could you give me some literature references, please.

 Notice that this requires a formal (i.e. logical) treated

-OK, but not only "this" requires a formal (i.e. logical) treated.
Therefore, I think that in the near future, commercially available intelligent software agents must be - "Ethically guaranteed" according to some ISO XXXXX standards with a specification of the name of an ethics used  - and this information shoud always be given to its end-user, too.

With my best wishes,

--- Adam

A REMARK after 7 years: the problem of "ethical agents" has been analized in a software word "Information Ecosystem"  of   "infohabitants" in the EU ALFEBIITE Project (2000-2003), but the results seem to be far yet to the practical conclusions. Emergent Societies of Information Agents.

 - In general, the problem of the ethics of intelligent software agents is open. 
 More, nowadays, it has a new dimension related to the possible behaviors of  new generation of mobil 
 Discussion List:  Your comments  to Absintelagent e-group  will be very appreciated (new Link of AIA e-group).

Meta-Ethics Comments: TOGA  proposal of the updating of the definition of ethics 

Ethics (human):
a set of top normative rules and values related to interpersonal interactions, evolved from tradition and cultural consensus of a society, and autonomously managed and applied by individuals in their decisions (a moral choice).

Ethics (artificial intelligent systems/agents):
a system of top normative rules and values derived from a chosen human ethics in an explicity distinguished domain of intelligent system/agent activity. It is implemented as a set of top level preferences, and autonomously used for decision-making in its interaction with other intelligent entities in this domain.

Social Ethics (unified socio-cognitive definition):
a system of top normative rules and values accepted by a community/society/group of intelligent entities (= a social entity) and appled by its members to the regulation of their reciprocal interactions.

Individual Ethics (unified socio-cognitive definition)
an ethical system (= a system of top normative rules and values regulating interactions
( intelligent entity - intelligent entity ) )  used only by an individual versus other members of its/his/her social entity.

Ethical/moral choice is a component of the generic scenario of  every cognitive decision-making. It is based on the top-down application of  highest/most general rules ( preferences rules) mandatory for every their specialization. 

Remark: According to the TOGA conceptualization framework, 

Ethics is a property of and refers only to self-active intelligent entities

(i.e. intelligent human/artificial agents). 

See: MEFIT Proposal -an expression of interest.

Meta-ethics (from Internet Encyclopedia of Philosophy:  http://www.utm.edu/research/iep/e/ethics.htm#Metaethics
The term "meta" means after or beyond, and, consequently, the notion of metaethics involves a removed, or bird's eye view of the entire project of ethics. We may define metaethics as the study of the origin and meaning of ethical concepts. When compared to normative ethics and applied ethics, the field of metaethics is the least precisely defined area of moral philosophy. Two issues, though, are prominent: (1) metaphysical issues concerning whether morality exists independently of humans, and (2) psychological issues concerning the underlying mental basis of our moral judgments and conduct. 

Meta-ethics is a study/research domain of the common properties of various existing and possible ethical systems from the perspective of their socio-cognitive values. (A.M.Gadomski, 2004).


From the meta-ethics perspective we can distinguish many different types of ethics. Taxonomy of ethics may depend on different factors but, most frequently, it is related to different religions,  ideologies, or group of people with common interests.

Basic two concepts of every ethics are: right/good and wrong/evil

For the reason of normative character of every ethics, represented mainly by rules and consensus-based importance indicators, the concepts of  wrong, evil refer to behaviours/actions which are against or not follow them.

For more  > Computational Axiology  a conceptualization framework

Seventh ETHICOMP International Conference on the Social and Ethical Impacts of Information and Communication Technologies. Hosted in the University of the Aegean, Syros, Greece, 14 to 16 April 2004. http://www.ccsr.cse.dmu.ac.uk/conferences/ethicomp2004/
Google  25/03/03   "intelligent agent ethics" - only MKE's 2 docs,   "agent's ethics" - 27 doc  , 
                                               "agent ethics" - 207,   meta-ethics  - 2980. 
                           21/10/03     "computational ethics"  - 34 (3,4 from MKS), 
                                               "computational axiology" - 3 (1,2 from MKS).
                           03/08/04    : "business of ethics" -  598.
                           23/09/04   :  "business of ethics"  - 754, "agent ethics" - 368 (1st MKS).

        MKS (Meta-Knowledge Server)    |    AGENTS and INTELLIGENCE   HID Research    |    A.M.Gadomski
For citation: A.M.Gadomski,About Intelligent Agent Ethics and ... Computational Axiology : a Socio-Cognitive Perspective,
Meta-Knowledge Engineering Server, ENEA, http://erg4146.casaccia.enea.it/Ethic-Agents.html,  Last updating: 23 Sep.2004.

© 1996-2003, ENEA. A.M.Gadomski. All rights reserved. No permission is granted to download and save professional images, code, or any other material from these pages other than for viewing and  citation purposes.  These are research pages representing the opinions of the contributors, but not necessarily of ENEA. The information is provided without any warranty. It is intended to be useful, but is certainly not always complete nor up-to-date.  Last updating: 21 Oct. 2003, last edited 23 Sep.2004.