5.10 Bias and Unethical use of Artificial Intelligence

Introduction

Artificial Intelligence has become an integral part of modern society. It has brought positive impacts in many areas of our daily lives but has some negative impacts as well. Biased and unethical use of AI-based systems has become a big concern all over the world as it can result in injustice to specific people.

Definition

Bias means unfairness or discrimination against someone based on race, ethnicity, gender, age, etc.

Unethical means something that is not morally correct, such as unfair usage of AI related to violation of personal data or gaining unauthorized access to other's systems.

Main Idea

AI algorithms are developed by humans who can intentionally or unintentionally introduce bias in them. AI-based systems are not thoroughly checked for racial or gender biases. If the data used in training an AI system is incorrect, impartial, incomplete, or biased in any way, then the resulting AI system can cause injustice, hurt people, and cause harm.

Core Concept

The main types of bias and unethical use of AI are:

• Privacy Breaches
• Social Manipulation and Misinformation
• Cyber Security Threats
• Financial Market Manipulation
• Deepfake Videos

Detailed Explanation

Real-World Examples of AI Bias

Amazon Hiring Algorithm (2015): In 2015, Amazon realized that their AI algorithm used for hiring employees favored men over women. The reason for this injustice was that the algorithm was based on the number of resumes submitted over the past ten years, which mostly belonged to male applicants.

Facial Recognition Technology: It was observed that facial recognition technology has sometimes produced wrong results for people with darker skin tones. This can cause injustice to people who are misidentified by the system in a criminal investigation.

Privacy Breaches

Privacy breaches are concerned with the violation of protection of personal data. AI systems have the ability to store and process huge amounts of data which is vulnerable to privacy violation.

Example: AI systems collecting personal data without consent or using it for unauthorized purposes.

Social Manipulation and Misinformation

AI systems can spread propaganda and manipulate social media trends such as people's likes and dislikes or changing their opinions related with all aspects of society.

Example: Bots spreading fake news or manipulating public opinion during elections.

Cyber Security Threats

AI systems can employ techniques that can be used in cyber-attacks to gain unauthorized access to other's computer systems. This can lead to financial loss, privacy invasion, or identity theft.

Example: AI-powered malware that adapts to bypass security measures.

Financial Market Manipulation

AI systems can be used to manipulate the stock market, which can lead to economic instability.

Example: High-frequency trading algorithms that manipulate stock prices.

Deepfake Videos

Deepfake videos are computer-created AI-based realistic fake videos. These videos misrepresent someone or say something that was not actually done or said. Such videos can spread misinformation, defamation, or change public opinion.

Example: Fake videos of politicians saying things they never said.

Intermediate Level

At the intermediate level, you should understand the causes and consequences of AI bias.

Causes of AI Bias:

Consequences of AI Bias:

Important Points

• - Bias means unfairness or discrimination.
• - Unethical means not morally correct.
• - AI algorithms are developed by humans who can introduce bias.
• - Amazon hiring algorithm favored men over women.
• - Facial recognition has produced wrong results for darker skin tones.
• - Types: privacy breaches, social manipulation, cyber security threats, financial market manipulation, deepfake videos.
• - Biased data leads to biased AI.
• - AI bias can cause injustice, harm, and loss of trust.

Remember

• - Bias = unfairness.
• - Unethical = not morally correct.
• - AI bias comes from biased data.
• - Amazon hiring = gender bias.
• - Facial recognition = racial bias.
• - Types: privacy, social manipulation, cyber threats, financial manipulation, deepfakes.
• - Biased data → biased AI.
• - Consequences: injustice, harm, loss of trust.

Common Mistakes

• - Thinking that AI is always objective and unbiased (it is not).
• - Assuming that AI bias is always intentional (it can be unintentional).
• - Forgetting that biased data leads to biased AI.
• - Thinking that AI bias only affects certain groups (it affects everyone).
• - Ignoring the ethical implications of AI.
• - Assuming that AI bias is easy to fix (it is complex).
• - Thinking that AI bias is only a technical problem (it is also social and ethical).

Quick Revision

• Bias: unfairness or discrimination.
• Unethical: not morally correct.
• AI algorithms developed by humans can introduce bias.
• Amazon hiring algorithm favored men over women.
• Facial recognition produced wrong results for darker skin tones.
• Types: privacy breaches, social manipulation, cyber security threats, financial market manipulation, deepfake videos.
• Biased data leads to biased AI.
• Consequences: injustice, harm, loss of trust.
• Solutions: diverse data, bias detection, fairness metrics, explainable AI, ethical guidelines, regulation.

60-Second Revision

• Bias = unfairness.
• Unethical = not morally correct.
• AI bias comes from biased data.
• Amazon hiring = gender bias.
• Facial recognition = racial bias.
• Types: privacy, social manipulation, cyber threats, financial manipulation, deepfakes.
• Biased data → biased AI.
• Consequences: injustice, harm, loss of trust.
• Warning: AI is not automatically fair; it reflects the biases in its training data.