Machine Ethics: A Brief Tutorial
Jim Moor
Philosophy
Dartmouth College
Hanover, NH 03755, USA
[email protected]
Science is Value-Laded
Computer Science is no exception
Normativity in Science
Values and norms are an essential part of all
productive sciences qua sciences.
They are used not only to establish the
suitability of existing claims but also to select
new goals to pursue.
Scientific evidence and theories are often
evaluated as either good or bad and scientific
procedures as what ought or ought not be done
Historical computer science illustration of such evaluation:
Herbert Simon’s reply to Jacques Berleur
November 20, 1999
“My reply to you last evening left my mind nagged by the question
of why Trench Moore, in his thesis, placed so much emphasis on
modal logics. The answer, which I thought might interest you,
came to me when I awoke this morning. Viewed from a computing
standpoint (i.e., discovery of proofs rather than verification), a
standard logic is an indeterminate algorithm: it tells you what you
MAY legally do, but not what you OUGHT to do to find a proof.
Moore viewed his task as building a modal logic of “oughts” -- a
strategy for search -- on top of the standard logic of verification.”
Normativity in Science
Moreover ethical norms often play a role in
the evaluation of science done properly. This
is particularly true as a science becomes
more applied.
Welcome to philosophy!
Of course, computer science has had
philosophical roots from the beginning.
Consider Hobbes or Pascal or Leibniz
Computer Ethics vs. Machine Ethics
Roughly, Computer Ethics emphasizes the
responsibility of computer users to be ethical,
for example with regard to privacy, property,
and power.
Whereas
Machine Ethics emphasizes building ethical
abilities and sensitivities into computers
themselves.
Grades of Machine Ethics
Normative Computer Agents
Ethical Impact Agents
Implicit Ethical Agents
Explicit Ethical Agents
Autonomous Explicit Ethical Agents
Full Ethical Agents
Machine Ethics
Normative Computer Agents
Ethical Impact Agents
Implicit Ethical Agents
Explicit Ethical Agents
Autonomous Explicit Ethical Agents
Full Ethical Agents
Where is the machine
decision-maker?
BigBelly
Compacts 300 gallons of
garbage. In photo in NYC.
Solar powered
Wireless sensor transmits for
pickup when nearly full
http://www.wired.com/news/planet/0,2782,66993,00.html
Machine Ethics
Normative Computer Agents
Ethical Impact Agents
Implicit Ethical Agents
Explicit Ethical Agents
Autonomous Explicit Ethical Agents
Full Ethical Agents
Robot Camel Jockeys of Qatar
“Camel jockey robots, about 2 feet high, with a right hand to bear the whip and a left
hand to pull the reins. Thirty-five pounds of aluminum and plastic, a 400-MHz
processor running Linux and communicating at 2.4 GHz; GPS-enabled, heart ratemonitoring (the camel's heart, that is) robots.”
http://www.wired.com/wired/archive/13.11/camel.html?tw=wn_tophead_4
Robot Camel Jockeys of Qatar
“Every robot camel jockey bopping along on its improbable mount
means one Sudanese boy freed from slavery and sent home.”
http://www.wired.com/wired/archive/13.11/camel.html?tw=wn_tophead_4
Ethical Impact Agents:
How well might machines themselves handle basic ethical issues of
privacy, property, power, etc.
Machine Ethics
Normative Computer Agents
Ethical Impact Agents
Implicit Ethical Agents
Explicit Ethical Agents
Autonomous Explicit Ethical Agents
Full Ethical Agents
Implicit Ethical Agents:
Ethical considerations such as safety
and reliability built into the machine.
Examples:
ATM
Air Traffic Control Software
Drug Interaction Software
Machine Ethics
Normative Computer Agents
Ethical Impact Agents
Implicit Ethical Agents
Explicit Ethical Agents
Autonomous Explicit Ethical Agents
Full Ethical Agents
Explicit Ethical Agents:
Ethical Concepts Represented and Used
Example: Carebots
Isaac Asimov’s Three Laws of Robotics
(1) A robot may not injure a human being, or,
through inaction, allow a human being to come to
harm.
(2) A robot must obey the orders given it
by human beings except where such orders
would conflict with the First Law.
(3) A robot must protect its own existence as long
as such protection does not conflict with the
First or Second Law.
Just Consequentialism
Core Values
Consequences (foreseeable)
Justice
1. Rights and Duties
2. Impartiality in Policy
Core Values
Life
Happiness
Ability
Security
Knowledge
Freedom
Opportunities
Resources
A Way to Remember Core Values
Happy Life?
Ability
Security
Knowledge
Freedom
Opportunities
Resources
Test Question for Judging Policies:
Is this a policy that a fully informed, rational,
impartial person would freely advocate as a
public policy?
A few observations about just consequentialism
Bounded rationality
Not maximizing
Impartiality
Dynamic and revisable policies
Handles conflicts
Procedure not algorithm
Often not just one correct solution
Could a machine use
Just Consequentialism?
Machine Ethics
Normative Computer Agents
Ethical Impact Agents
Implicit Ethical Agents
Explicit Ethical Agents
Autonomous Explicit Ethical Agents
Full Ethical Agents
Autonomous Ethical Agents:
Make ethical decisions and actions (not
merely decisions and actions that are
ethical) on a dynamic basis as interact
with environment. These agents are not
conscious but are not merely
responding with canned responses to
situations. They may learn and adapt
their behavior.
Examples?
Autonomous Ethical Agents:
Make ethical decisions and actions (not
merely decisions and actions that are
ethical) on a dynamic basis as interact
with environment. These agents are not
conscious but are not merely
responding with canned responses to
situations. They may learn and adapt
their behavior.
What would an example be like?
Disaster Relief Software Agent
Suppose it receives information about who is in
need, how badly they are injured, where
supplies are, etc. Makes decisions with limited
resources about who gets what. Triage
sometimes required.
Might it not run FEMA better than humans?
More Ethically?
More Ethical Military Machines?
Driverless Cars
Driverless Cars
Driverless Truck
The Trolley Problem
Why is Machine Ethics Important?
Normative Computer Agents
Ethical Impact Agents
Implicit Ethical Agents
Explicit Ethical Agents
Autonomous Explicit Ethical Agents
Full Ethical Agents
Why is Machine Ethics Important?
1. Ethics is important
2. Machines will have increased control and autonomy
3. Opportunity to understand ethics better
Can a Machine Ever Be A Full Ethical?
Philosophical Objections to Machine Ethics
Type 1: Ethics cannot exist above the bright red line.
Type 2: Machines can’t exist below it.
Normative Computer Agents
Ethical Impact Agents
Implicit Ethical Agents
Explicit Ethical Agents
Autonomous Explicit Ethical Agents
Full Ethical Agents
Philosophical Objections to Machine Ethics
Type 1: Ethics cannot exist above the bright line.
Ethics has no basis
Philosophical Objections to Machine Ethics
Type 1: Ethics cannot exist above the bright line.
Ethics has a basis, but machines can’t do it
Philosophical Objections to Machine Ethics
Type 1: Ethics cannot exist above the bright line.
Ethics has a basis, machines can do it, but
should not.
Joseph Weizenbaum continues...
“Computers can make judicial decisions, computers
can make psychiatric judgments. They can flip
coins in much more sophisticated ways than can the
most patient human being. The point is that they
ought not to be given such tasks. They may even be
able to arrive at “correct” decisions in some cases –
but always and necessarily on bases no human being
would be willing to accept.”
Computer Power and Human Reason, p. 227
Philosophical Objections to Machine Ethics
Type 1: Ethics cannot exist above the bright line.
Ethics has a basis, machines can approximate it,
but machines above the bright line will lack a
crucial ingredient for real ethical agents:
Intentionality, Consciousness, Free Will,
Responsibility,….
Normative Computer Agents
Ethical Impact Agents
Implicit Ethical Agents
Explicit Ethical Agents
Autonomous Explicit Ethical Agents
Full Ethical Agents
Philosophical Objections to Machine Ethics
Type 1: Ethics cannot exist above the bright line.
Type 2: Machines can’t exist below it
Open question ?