ERS 10-11

10

10

Definition of fairness

the absence of any prejudice or favoritism toward an individual or group.

Discrimination and types

refers to misdesigned algorithms or biased outcomes that unfairly favor or disadvantage specific groupsbasedonattributes such as race, gender, or age.

Direct discrimination
(decisions are explicitly influenced by sensitive attributes)

Indirect discrimination
(decisions use apparently neutral attributes that are correlated with sensitive ones)

Systemic discrimination
(Unfairness built into institutions or broader systems)

Statistical discrimination
(Judging people by group averages rather than individual qualities)

Summary. Bias, Fairness and Discrimination

Equality VS Equity

Fairness measurement

Unawareness
The system tries to be fair by removing sensitive attributes like gender or race from the data.

Demographic Parity
The rate of positive outcomes should be the same across groups.

Equalized Odds
The system should have equal true positive and false positive/false negative behavior across groups.

Predictive Rate Parity
The probability that a positive prediction is correct should be the same across groups.

Individual Fairness
Similar individuals should be treated similarly by the model.

Counterfactual Fairness
A decision is fair if it would stay the same even if a sensitive attribute were changed in a hypothetical scenario.

Example of calculations

Definition and types of transparency

understanding how AI/robot systems make decisions, generate outputs, and use data.

Algorithmic transparency
(Internal workings of the system: the logic, processes, algorithms, and how data is used to make decisions)

Interaction transparency
(Communication between users and the AI/robot)

Social transparency
(Concerns the broader societal impact of AI, including ethical effects, public behavior, cultural values, social norms, and regulation)

Definition of accountability and reasons why we need it

Responsibility and answerability for actions and outcomes, including the duty to justify them to an authority.

Legal compliance
(Accountability helps ensure that robot systems follow existing laws and regulations)

Liability and compensation
(If damage or loss occurs, accountability helps determine who is responsible and who pays for damages)

Error and malfunction management
(It helps identify faults, unintended behaviors, and system failures, which improves reliability and safety)

Connection to transparency
(AI developers accountable requires understanding how the system works, so transparency supports accountability)

11

11

Types of Human-Robot interactions

Coexistence
(Humans and robots divided by a line)

• Space: Separated

• Tasks: Unrelated

• Contact: Impossible

• Context Awareness: Not necessary

Cooperation
(Humans and robots work in a shared space, but their tasks are linked rather than fully shared)

• Space: Shared

• Tasks: Linked

• Contact: Possible

• Context Awareness: Minimal

Collaboration
(Humans and robots work together in the same shared space on the same shared tasks, which requires stronger interaction)

• Space: Shared

• Tasks: Shared

• Contact: Frequent and necessary

• Context Awareness: High

Challenges in human-robot interaction

Cobot design
(Collaborative robots must be designed so they can work safely and efficiently with humans)

Safety and security
(Robots should operate without creating risks for human workers, so ensuring safe and secure operation is a central challenge)

Interaction quality
(Robots need to understand and predict human actions, intentions, and emotions to support good collaboration)

Explain a uncanny valley effect

robot becomes more human-like, people usually like it more at first, but if it becomes almost human, but not fully, it can suddenly feel strange, unsettling, or creepy

Design principles for robots

Human-centered design
(Design robots around people’s needs, context, and motivations using steps like empathize, define, ideate, prototype, test)

Usability
(A robot should not only work technically, but also be easy and useful to use)

Match appearance to capability
(A robot should not seem more intelligent, emotional, or capable than it really is)

Support safe interaction
(Robots should be designed for safe, efficient, and intuitive collaboration with humans)

What is trust calibration in HRI?

matching human trust to the robot’s real capabilities so people rely on it appropriately, safely, and effectively

Summary

12

12

What are the impacts of robots on society?

What is privacy and why is it important?

What regulations are existing that prevent companies in misusing our data?