Note: This article is present in its copy on author’s medium.com account as well.

#ai #artificialintelligence #robotics #machinelearning

This is a summary of a chapter from the latest book on AI Ethics, by C. Bartneck et al. (2021) and has been self-written.

Ethics define the moral set of rules that define a person’s decisions. What is right what is wrong, and what is morally correct and what is morally wrong. Ethics in AI refers to what an AI machine or robot needs to do to be morally right and what rules it needs to follow.

Normative ethics

It’s a general truth valid for everyone for example a crime is a crime whether performed by a human or by a robot. There are the following kinds of Normative ethics:

Kinds of Ethics

1. Deontological Ethics: These are concerned with duty and obligations. Corporate Social Responsibility (CRS) comes under this framework. The main aim is motivation behind the responsibility. However, action and results do play a secondary role in this. Action and intention are both important.

2. Consequentialist ethics: Here the aim is predictable consequences that an act has. CRS above plays no role in this; only defined outcomes are measured here.

3. Virtue Ethics: The aim here is to follow virtues such as wisdom, justice, fortitude, etc. to attain the goals. Acting using the virtues of both the provider and those influenced by the actions and outcomes is a key consideration here.

Meta-ethics

It is the theory of normative ethics. It is a theory of normative ethics and consists of the following:

1. Existence: What objects in the world under consideration exist and have moral features?

2. Meaning or semantics: Associates the assigning of terms good, bad, right, or wrong with objects in the world.

3. Knowledge: This corresponds to how to know moral truth.

Applied Truth

Here ethical rules are passed. This can be explained by taking the example of ethical medications, bioethics, and even business ethics, to mention a few. This term applied ethics or truth means different things under different situations and hence is a relative terminology.

Relationship Between Ethics and Law

As per the author(s) ethics start where the law ends. Real-life entities would then have legal duties and ethical duties. The authors say, that some legal duties have ethical duties as well, for example, cars causing more pollution is both a legal and an ethical issue now. This makes ethics as a subject of soft law while legal rules are the hard laws.

Machine Ethics

The aim of the subject of Machine Ethics is to determine what is needed to make an ethical AI that makes moral decisions. The key difference is that humans have feelings while machines do not have feelings, though these can be added to a machine. The goal to make machines that can take moral decisions is not a cakewalk. An example is the robot Sophia which is considered to be a cooperate publicity stunt.

Examples

The aim to create machines with morals and feelings starts with sensors and input trackers. These inputs can go to Robot tracking systems. For example, the authors provide this example, that a car was over speeding, and following the rules of logic embedded in the robot, the robot makes a ticket for the car driver. However, other events were happening at the same time. And the robots lack moral decision-making to decide what task to perform first since the computation of logic to show a ticket to the car driver can take time. Here, the author suggests assigning utility numbers to all things the robot is working on and can refer to utility numbers to compute what next to do to avoid a clash of duties.

Moral Diversity and Testing

Testing and certifying that an AI is correct and ethical is a prime area of research these days.

AI fairness 360 open-source toolkit and audit-AI provides the certificate if the software has a bias. Other companies such as O’Neil Risk Consulting and Algorithmic Auditing, are working on testing for Fairness Flow, and to find if there is bias in the AI products.

Reference

C. Bartneck et al., An Introduction to Ethics in Robotics and AI, SpringerBriefs in Ethics, 2021, https://doi.org/10.1007/978-3-030-51110-4_2