Chapter 4 — Ai Ethics And Societal Impact

What is AI Ethics?

AI Ethics is the study of moral principles and rules that guide the design, development, deployment and use of artificial intelligence systems. It helps ensure AI systems are fair, safe, transparent and respect human rights.

Why it matters

AI systems make decisions that affect people’s lives—recommending jobs, deciding loan approvals, recognizing faces, or driving cars. If these systems are biased, unsafe or secret, they can harm individuals or groups. Ethics helps prevent harm and build trust.

Core principles (simple explanations)

• Fairness – AI should not discriminate against people because of race, gender, religion or other protected traits.
• Transparency – It should be possible to understand how an AI reached its decision (explainability).
• Accountability – People and organizations should be responsible for AI outcomes.
• Privacy – Personal data used by AI must be protected and used with consent.
• Safety and robustness – AI should perform reliably and safely in expected situations and handle unexpected inputs gracefully.
• Human oversight – Humans should be able to review, override or stop AI decisions when needed.

How ethical issues can arise (simple lifecycle view)

• Data collection – Biased or incomplete data can produce biased models.
• Model training – Algorithms may learn unfair patterns or amplify existing bias.
• Deployment – An AI used in a new context can behave in harmful ways.
• Monitoring – Without checks, problems may go unnoticed or get worse over time.

How to reduce ethical risks

• Use diverse and representative data sets; check for missing groups.
• Measure performance separately for different groups (e.g., by gender or age).
• Document choices (what data, why model chosen) so decisions are transparent.
• Include humans in the loop for important decisions (human review).
• Protect user data with privacy techniques and clear consent.
• Test systems in many conditions and monitor them after deployment.

Simple classroom activity idea

Give students a small dataset (e.g., loan approvals with applicant attributes). Ask them to split results by groups (e.g., by gender) and compute accuracy and other metrics to see if the system treats groups differently. Discuss how to improve fairness.

Takeaway

AI Ethics is about creating AI that respects people, reduces harm, and is trustworthy. It combines technical checks (metrics, tests) with social choices (rules, responsibility).

What is privacy? Privacy is the right of a person to control who sees and uses their personal information. In the context of AI, privacy means protecting data about individuals (name, address, health, behaviour, location) from unauthorized access, misuse or unwanted disclosure.

What is data protection? Data protection are the technical, legal and organisational measures used to keep personal information safe and ensure it is processed fairly and lawfully. It includes how data is collected, stored, used, shared and deleted.

Why it matters

• Protects against identity theft, financial loss and harm (e.g., stalking, discrimination).
• Maintains trust between people and institutions (schools, apps, companies).
• Prevents biased or unfair decisions by AI systems built on poor or leaked data.

Types of data

• Personal data: information that can identify a person (name, phone number).
• Sensitive data: health, religion, biometrics, financial details — needs extra protection.
• Non-personal data: aggregated or anonymised data that cannot identify an individual.

Key principles of good data protection

• Consent: collect and use data only with clear permission.
• Purpose limitation: use data only for the stated reason.
• Data minimisation: collect only what is needed.
• Transparency: tell people how their data will be used.
• Accuracy: keep data correct and up to date.
• Storage limitation: keep data only as long as necessary.
• Security: protect data with technical measures (encryption, access control).

Common protection methods

• Encryption — scrambles data so only holders of the key can read it.
• Anonymisation & pseudonymisation — remove or mask identifiers so individuals can’t be easily re-identified.
• Access controls — limit who can view or change data (passwords, roles).
• Secure transmission — use HTTPS and secure channels when sending data over the internet.
• Privacy by design — include privacy protections during system design, not after.

AI-specific concerns

• Training data may contain private information; models can unintentionally memorize and reveal it.
• Re-identification: combining datasets can reveal identities even if each dataset was anonymised.
• Bias & fairness: poor data protection can lead to datasets that reflect or amplify discrimination.

Practical advice for students

• Use strong, unique passwords and two-factor authentication.
• Check app privacy settings; limit what you share publicly on social media.
• Think before sharing sensitive details (health, location, identity documents).
• Delete old accounts and data you no longer use.

Summary: Privacy and data protection ensure that personal information is collected and used responsibly. They combine technical tools (encryption, anonymisation), principles (consent, minimisation) and laws/policies to protect individuals from harm and preserve trust in AI systems.

Bias in Artificial Intelligence means that a system consistently favors some outcomes or groups over others because of the data, the design of the algorithm, or the way the problem was defined. Bias can be unintentional and comes from incomplete, unrepresentative, or noisy data, or from human decisions during model building.

Fairness is the goal of making AI systems treat people and groups equitably. Fairness means similar people in similar situations should receive similar outcomes, and that sensitive attributes (like gender, caste, religion, race) should not cause unjust advantage or disadvantage.

Discrimination is when individuals or groups are treated differently in a harmful way because of a protected characteristic. In AI, discrimination can occur if a model’s decisions result in systematically worse outcomes for certain groups.

Causes and examples of bias: training data that under-represents a group (sample bias), historical prejudice embedded in records (historical bias), labels that reflect human prejudice (label bias), or model choices that prioritize one objective over fairness (algorithmic bias).

Consequences: unfair hiring, denial of loans, wrongful identification, unequal access to services, and reinforcement of social inequalities.

Ways to reduce bias and increase fairness: collect better and more representative data, include fairness-aware objectives during model training, test models separately for different groups, use techniques like re-sampling, re-weighting, or fairness constraints, and involve diverse teams and human oversight.

Simple classroom idea: split a dataset by group (e.g., boys/girls or regions), compute the model’s accuracy and error rates for each group, and discuss differences and possible causes.

Transparency means that the way an AI system works (its data, design and decision process) is open and understandable to people who use it or are affected by it. Explainability means the system can provide clear, simple reasons for a particular decision or prediction.

Why they matter: transparent and explainable AI builds trust, helps detect errors and bias, and makes it possible to hold systems and designers accountable. For students, parents and teachers, explainability means you can see why an AI gave a certain grade suggestion, recommended content, or flagged an answer.

How it is achieved (simple approaches):

• Use inherently interpretable models: linear models and decision trees show how input features affect the output directly.
• Provide post-hoc explanations: techniques like LIME or SHAP summarize how much each input influenced a single decision.
• Document data and code: publish what data was used, how it was cleaned, and what rules or objectives the system follows.
• Give human-centered explanations: present reasons in plain language (e.g., "Your loan was denied because income < required threshold and credit history low").

Trade-offs and limits: simpler models are easier to explain but sometimes less accurate. Complex models (deep neural nets) may be more accurate but harder to explain. Explainability techniques give approximations and may not capture every internal detail.

Practical advice: always ask for (1) what data was used, (2) what features matter most, and (3) a clear, testable reason for any decision that affects people.

Accountability and Responsibility

Responsibility is the duty to perform specific tasks or to ensure certain outcomes. In AI contexts it means the people or organizations who must design, test, deploy, maintain or supervise an AI system and follow ethical, legal and safety requirements.

Accountability is the obligation to explain and justify actions and to accept consequences when something goes wrong. Accountability requires transparency (records and explanations), mechanisms for review or audit, and enforcement or remedies if duties are not met.

Key differences (simple)

• Responsibility = who does the work or holds a duty.
• Accountability = who answers for the outcome and faces consequences.
• One actor can be responsible, another accountable; sometimes they are the same person or organization.

How this applies across the AI lifecycle

• Data collection: Responsibility—data collectors/annotators; Accountability—data controller/organization for data quality and consent.
• Model development: Responsibility—ML engineers, data scientists; Accountability—team lead or product owner for model behaviour and fairness.
• Testing and validation: Responsibility—QA/test team; Accountability—project manager or regulator for safety tests.
• Deployment and monitoring: Responsibility—operations team; Accountability—service owner for real-world impacts and remediation.

Practical mechanisms that make accountability real

• Documentation: model cards, data sheets, design logs that record choices and trade-offs.
• Traceability and logging: keep records of inputs, model versions, and decisions for audits.
• Human-in-the-loop and escalation: design processes that allow humans to intervene and reverse harmful actions.
• Redress and enforcement: clear complaint channels, incident investigation, and corrective measures.
• Clear role assignment: RACI or similar frameworks to state who is Responsible, Accountable, Consulted, and Informed.

Ethical and legal points

• Accountability helps ensure fairness, safety, and respect for rights (privacy, non-discrimination).
• Laws may require organizations to be accountable (data protection regulations, safety standards).
• Students should understand that ethical AI requires both people who act responsibly and systems that make those actors answerable.

Student summary: Responsibility is about duties; accountability is about being able to explain, justify and accept the consequences of how those duties were carried out. Both are needed to build trustworthy AI.

What it means: In AI, safety means designing systems that do not cause unintended harm and behave reliably even in unexpected situations. Security means protecting AI systems, their data, and their users from malicious attacks, theft, or misuse.

Key ideas

• Robustness — AI should handle noisy, unusual, or adversarial inputs without catastrophic failure (e.g., refuse to act if uncertain).
• Reliability & fail-safe — systems should degrade gracefully (fallback to human control or safe state).
• Privacy & confidentiality — protecting personal data used for training and inference (access controls, anonymization).
• Integrity & availability — ensure model code, parameters, and data are not tampered with and the service remains available.
• Transparency & explainability — users should get understandable reasons for important decisions so mistakes are detected and corrected.

Practical measures: Adversarial testing (checking how small input changes can fool the model), regular software patches, encryption of stored data, access control, monitoring for model drift, logging actions, differential privacy for sensitive datasets, and human-in-the-loop for high-risk decisions.

Why it matters for Class 9 students: Everyday AI (smartphones, school apps, social media, smart assistants) can make errors or expose data. Understanding safety and security helps students use and design AI responsibly.

What this topic covers
The social and economic impact of Artificial Intelligence (AI) looks at how AI changes people's lives, jobs, incomes, relationships, access to services, and the overall economy. It includes both positive effects (like increased productivity, new services, and better healthcare) and negative effects (like job displacement, widened inequality, and biased decision-making).

Key social impacts

• Access and inclusion: AI can increase access to education, health care, and government services through personalised learning, diagnostic tools and chatbots — but unequal access to devices and internet can create a digital divide.
• Privacy and surveillance: Widespread use of data-driven AI (camera analysis, tracking, profiling) can erode privacy and change social behaviour.
• Bias and fairness: Biased training data can cause AI systems to treat some groups unfairly (for example, facial recognition misidentifying people from certain ethnicities), leading to social injustice.
• Quality of life: AI assistants, accessibility tools (speech-to-text, predictive text), and smart devices can make daily life easier for many people.

Key economic impacts

• Productivity and growth: AI can automate repetitive tasks, make complex decisions faster, and boost output per worker, raising economic growth.
• Employment shifts: Some jobs and tasks may be automated away, while new roles (AI maintenance, data labeling, human-AI collaboration) are created. This causes sectoral shifts and requires reskilling.
• Income and inequality: Gains from AI may concentrate with capital owners and highly skilled workers, increasing income inequality unless policies redistribute benefits or provide training.
• Business models and markets: AI enables new services (personalised recommendations, dynamic pricing) and can strengthen network effects, sometimes leading to market concentration.

Balancing risks and benefits — common responses
Governments, schools and companies respond through reskilling/upskilling programs, ethical guidelines, fairness-aware design, data protection laws, taxation and social safety nets (e.g., unemployment support, targeted training). Inclusive design and public investment in connectivity reduce the digital divide.

Simple summary for students
AI changes how people work and live. It can make services better and economies richer, but also creates challenges like losing some jobs, unfair decisions, and unequal access. Understanding both sides helps society use AI responsibly.

What is Human-AI Interaction? Human-AI Interaction (HAI) is how people and artificial intelligence systems communicate, cooperate, and make decisions together. It includes interfaces (chatbots, voice assistants, recommender systems), collaboration with semi‑autonomous systems (robot assistants, adaptive learning platforms), and oversight of autonomous systems (self‑driving cars, automated loan decisions).

Why rights matter in HAI AI systems affect people's privacy, safety, opportunities and dignity. Rights and ethical principles guide how AI should be designed and used so that it respects people. Key rights and principles include:

• Privacy: personal information must be collected and used with consent and protection.
• Fairness and non‑discrimination: AI should not treat people unfairly because of gender, caste, religion, race or other protected attributes.
• Transparency and explainability: people should be able to understand or get a simple explanation of how an AI reached an important decision.
• Accountability: humans or organisations must be responsible for AI actions and harms.
• Safety and reliability: AI should be tested to avoid physical or emotional harm.
• Autonomy and consent: people should choose whether and how to interact with AI and should be able to opt out when possible.

Design approaches that protect rights include human‑in‑the‑loop systems (humans review important AI decisions), privacy‑by‑design (minimise data collection), explainable AI (simple explanations for users), and regular bias audits (check outcomes for different groups).

How students can apply this in daily life — ask questions when an app requests many permissions, check why a recommendation appears before relying on it, and know that automated decisions (school, banking, health) should offer human review. Educators should teach digital literacy: how data is collected, what consent means, and how to report harmful or biased outputs.

What it means: Responsible use and digital citizenship means using digital tools, the internet, and AI in ways that are safe, ethical, respectful, and lawful. A digital citizen understands their rights and responsibilities online, protects personal and others' privacy, respects intellectual property, and contributes positively to digital communities.

Key principles

• Respect: Treat others courteously; avoid hate speech, harassment, and cyberbullying.
• Safety: Protect personal data, use strong passwords and two-factor authentication, and avoid dangerous online interactions.
• Privacy: Share only necessary information; understand and manage privacy settings.
• Security: Recognize phishing, scams, and malware; keep software updated.
• Responsibility: Think before posting; consider long-term effects of online actions (digital footprint).
• Integrity and Attribution: Respect copyright and give credit for others' work; avoid plagiarism.
• Critical Thinking: Verify sources, evaluate information quality, and spot misinformation.
• Accountability: Accept consequences of online behavior and report abuse or harmful content.

Practical habits for students

• Think before posting: ask whether a post could harm someone or your future opportunities.
• Manage privacy: review app permissions and social media privacy settings regularly.
• Use strong passwords and 2FA; do not share passwords.
• Verify information using trusted sources before sharing.
• Ask permission before using or sharing someone else’s photos, writing, or ideas.
• When using AI tools, verify outputs, cite the tool when appropriate, and do not pass off generated content as solely your original work.
• Report cyberbullying, suspicious contacts, or illegal content to a trusted adult or platform moderators.

Consequences of irresponsible use: Loss of privacy or reputation, academic or legal penalties, identity theft, emotional harm to others, spread of misinformation, and long-term digital records that affect future education or jobs.

Classroom role: Teachers and schools set policies (acceptable use, anti-bullying, citation rules) and model good digital citizenship through lessons, discussions, role-play, and projects that teach evaluation of sources, privacy settings, and ethical use of AI.

What this topic means

When we build and use Artificial Intelligence (AI), we must follow rules and good practices so AI is safe, fair and respects people’s rights. These rules come from three related but different sources:

• Legal frameworks – Laws passed by governments that are binding and enforceable (breaking them can lead to fines or court cases). Example: data protection laws.
• Policy frameworks – Official guidelines or plans made by governments or organisations to shape how AI should be developed and used. They are not always laws but guide decisions and create standards.
• Ethical frameworks – Moral principles and values (like fairness, transparency and responsibility) that developers, companies and societies follow to make AI systems trustworthy.

Key ethical principles for AI

• Fairness: AI should not discriminate against people based on race, gender, caste, religion or other protected traits.
• Transparency & Explainability: People should be able to understand how important AI decisions are made.
• Privacy: Personal data must be collected and used with consent and protection.
• Accountability: Someone (developers, companies or operators) must be responsible when AI causes harm.
• Safety & Reliability: AI should work safely and as intended under normal and unexpected situations.
• Human oversight: Humans must be able to review or override important AI decisions.

How the three frameworks work together

Laws set the minimum rules. Policies translate laws into practise or fill gaps when laws are missing. Ethics guide behaviour beyond what the law requires—helping organisations do what is right even if it is not yet mandatory.

Who is involved?

Governments (make laws and national policies), companies (create internal policies and follow laws), researchers (create ethical guidelines and technical tools), civil society (watchdogs and public feedback), and users (people affected by AI).

Examples of legal & policy instruments

• GDPR (European data protection law) — strong legal rules about personal data, consent and the right to explanation.
• National laws and bills (for example, India’s data protection initiatives and IT rules) — set local requirements.
• OECD, UNESCO and IEEE AI guidelines — policy and ethical recommendations followed internationally.

Practical steps for safe AI (simple checklist)

• Identify risks (who may be harmed and how).
• Use minimal personal data and get clear consent.
• Test AI for bias and accuracy across groups.
• Document how the AI works and keep logs for audits.
• Allow human review and provide explanations for decisions that affect people.

Why it matters for students

AI is already part of many apps and services. Understanding these frameworks helps you know your rights (privacy, fairness), why some apps ask for permissions, and how to build or evaluate AI responsibly in the future.

AI for Social Good

"AI for Social Good" means using artificial intelligence to solve real-world problems that improve people’s lives and protect the planet. The focus is on health, education, environment, disaster response, accessibility and fair public services. When applied responsibly, AI can assist humans to make faster, fairer and more informed decisions.

Key goals

• Increase public welfare (better health, safety, education).
• Make services more accessible and inclusive.
• Protect environment and wildlife.
• Support quick responses during disasters and emergencies.

Typical AI workflow

1. Problem definition (what social problem to solve).
2. Data collection and cleaning (ensure quality and privacy).
3. Model selection and training (choose an appropriate algorithm).
4. Evaluation (accuracy, fairness, robustness).
5. Deployment with human oversight (human-in-the-loop).
6. Monitoring and update (watch for bias, drift, unintended consequences).

Ethical principles

• Benefit: prioritize positive social impact.
• Fairness: avoid discrimination across groups.
• Privacy: protect personal data and follow consent rules.
• Transparency: explain decisions to users when possible.
• Accountability: humans remain responsible for outcomes.

Case studies (short summaries)

• Healthcare — diabetic retinopathy screening: AI models analyze retina images to detect early signs of eye disease, enabling faster referrals to specialists and preventing vision loss. Ethics: ensure tested across different patient groups.
• Disaster response — satellite imagery analysis: After floods or earthquakes, AI scans satellite images to identify damaged buildings and blocked roads so responders prioritize aid. Ethics: timely updates and validation with local teams.
• Agriculture — crop disease detection: Mobile apps let farmers take photos of leaves; AI identifies disease and suggests treatments, increasing yields and reducing food waste.
• Environment — anti-poaching and wildlife monitoring: Acoustic sensors and camera traps analyzed by AI detect poachers or endangered animals, helping rangers act faster.
• Accessibility — speech-to-text and image descriptions: AI converts speech to text for the deaf and generates image descriptions for the visually impaired, increasing access to information.

Benefits and limitations

Benefits include speed, scale and improved decision support. Limitations: biased data can cause unfair outcomes, privacy breaches are possible, models can be overconfident, and real-world deployment needs human oversight.

Best practices for safe deployment

• Use diverse, representative data and test models across groups.
• Keep humans in the decision loop for critical actions.
• Document data sources and model behavior (transparency).
• Protect user data and use anonymization when possible.
• Monitor models in the field and update regularly.

In class, you can discuss one local problem (for example, helping a nearby clinic or designing a classroom accessibility tool) and sketch a simple AI solution following the workflow above. Emphasize that technology must be combined with ethics and community involvement to truly be "for social good."

Overview: This topic examines the main challenges society faces because of AI, promising future directions to make AI safer and fairer, and practical roles students can play to shape responsible AI use.

Key Challenges:

• Bias and fairness: AI systems trained on biased data can produce unfair outcomes for certain groups (race, gender, income). This reduces trust and can cause real harm.
• Privacy and surveillance: AI can collect, combine, and infer sensitive information from many data sources, threatening personal privacy.
• Transparency and explainability: Many AI models (especially deep learning) are complex and hard to interpret, making it difficult to understand or justify decisions.
• Accountability and governance: It is often unclear who is responsible when an AI system makes a harmful decision—the developer, deployer, or data owner?
• Safety and robustness: AI can fail in unexpected ways when faced with unusual inputs (adversarial examples, sensor errors), causing dangerous outcomes in critical systems.
• Job displacement and economic impact: Automation may change the kinds of jobs available and require new skills.
• Misuse and malicious uses: AI can be used to create deepfakes, amplify misinformation, automate cyberattacks, or design harmful biological agents.
• Digital divide: Unequal access to AI tools and education can widen social and economic inequalities.

Future Directions (how problems are being addressed):

• Explainable AI (XAI): Methods and tools that make model predictions understandable to humans so users can trust and verify decisions.
• Fairness-aware learning: Algorithms and training procedures designed to reduce bias in outcomes (pre-processing, in-processing, post-processing methods).
• Privacy-preserving techniques: Approaches such as federated learning, differential privacy, and homomorphic encryption that allow useful AI while protecting personal data.
• Robustness and safety engineering: Testing, monitoring, adversarial defenses and formal verification for critical AI systems (e.g., autonomous vehicles, medical devices).
• Regulation and ethical frameworks: Laws, standards and industry guidelines to govern responsible AI development and deployment.
• Human-AI collaboration: Interfaces and workflows that keep humans in the loop for oversight, final decisions, and value judgments.
• Multidisciplinary education and research: Combining computer science with ethics, law, social sciences and domain expertise to build socially-aware AI.

Student Role (what students can do now):

• Learn basics of AI and ethics: Understand how AI works, its limits, and ethical concepts like fairness, privacy and consent.
• Practice data responsibility: Learn about data collection, cleaning, consent and anonymization—use only ethical datasets in projects.
• Ask critical questions: Question how datasets were collected, who benefits, and who may be harmed by an AI system.
• Build explainability into projects: Use simple, interpretable models when appropriate and document decisions and limitations.
• Share and collaborate: Participate in school projects, hackathons and community initiatives to create beneficial AI solutions (e.g., accessibility tools).
• Advocate and educate: Raise awareness among peers and family about safe AI use, privacy settings, and misinformation.
• Follow continuous learning: Acquire coding, statistics and critical-thinking skills; learn about new privacy-preserving tools and regulations.
• Report misuse: Speak up if you see biased outputs, privacy breaches or harmful uses of AI in your community.

Summary: The challenges around AI are technical, social and legal. Future directions focus on fairness, privacy, transparency, safety and governance. Students can play a meaningful role by learning, building responsibly, asking critical questions, and advocating for ethical AI.