Unit 1: Revisiting AI Project Cycle & Ethical Frameworks for AI
This unit revisits how an AI project is planned and completed through the AI Project Cycle. It also explains AI domains and shows how ethical frameworks help AI solutions avoid unintended harm.
AI Project Cycle
The AI Project Cycle is the cyclical process followed to complete an AI project. It gives a proper framework for moving towards the goal of an AI project. Like planning a greeting card, an AI project also needs clear steps before it reaches the final stage.
The six stages of the AI Project Cycle are:
- Problem Scoping
- Data Acquisition
- Data Exploration
- Modeling
- Evaluation
- Deployment
Problem Scoping is the first stage. In this stage, the goal of the AI project is set by clearly stating the problem to be solved. The different parameters that affect the problem are studied so that the problem becomes clearer. Students fill in the 4Ws problem canvas and write a problem statement.
Data Acquisition comes next. Data becomes the base of the project. It helps in understanding the parameters connected to the problem. Data is collected from various reliable and authentic sources.
Data Exploration is done after data is collected. Since the data may be large in quantity, it can be shown through visual representations. These include graphs, databases, flow charts, maps, and similar forms. This makes it easier to understand the patterns in the acquired data.
Modeling means deciding the type of model that can help achieve the goal. Different models can be researched and selected if they give suitable output. The selected models can be tested to find the most efficient one. The most efficient model becomes the base of the AI project, and the algorithm is developed around it.
Evaluation happens after modeling is complete. The model is tested on newly fetched data. The results help in evaluating the model and improving it.
Deployment is the final stage. It is important for successful integration and operation of AI solutions in real-world environments. It helps the solution deliver value and impact to users and stakeholders.
Introduction to AI Domains
Artificial Intelligence becomes intelligent according to the training it gets. For training, the machine is given datasets. The data changes according to the application for which the AI algorithm is being developed.
With respect to the type of data fed into the AI model, AI models are broadly categorized into three domains:
- Statistical Data
- Computer Vision
- Natural Language Processing
Statistical Data is related to data systems and processes. In this domain, the system collects a lot of data, maintains datasets, and derives meaning from them. The information extracted through statistical data can be used to make decisions. Price comparison websites are examples of this domain. These websites use large amounts of data to compare the price of a product from multiple vendors in one place. Examples include PriceGrabber, PriceRunner, Junglee, Shopzilla, and DealTime. Such websites can be found in domains such as technology, hospitality, automobiles, durables, apparel, and more.
Computer Vision, abbreviated as CV, is a domain of AI where a machine gets and analyses visual information and then predicts decisions based on it. The process includes image acquiring, screening, analysing, identifying, and extracting information. This helps computers understand visual content and act on it. Inputs to machines can include photographs, videos, pictures from thermal or infrared sensors, indicators, and other sources. Computer vision projects translate digital visual data into descriptions and then into computer-readable language to support decision-making. The main objective is to teach machines to collect information from pixels.
Examples of Computer Vision include agricultural monitoring and surveillance systems. In agriculture, computer vision is used for crop monitoring, pest detection, and yield estimation. Drones with cameras capture aerial images of farmland, which are analysed to assess crop health and improve farming practices. In surveillance systems, computer vision helps monitor public spaces, buildings, and borders. It can detect suspicious activities, track individuals or vehicles, and give real-time alerts to security personnel.
Natural Language Processing, abbreviated as NLP, is a branch of artificial intelligence that deals with interaction between computers and humans using natural language. Natural language means the language spoken and written by people. NLP uses algorithms to extract information from spoken and written words. Its main objective is to read, decipher, understand, and make sense of human languages in a valuable manner.
Examples of NLP include email filters and machine translation. Email filters are among the basic and initial online applications of NLP. Spam filters identify certain words or phrases that signal a spam message. NLP is also used in machine translation systems like Google Translate and Microsoft Translator to translate text from one language to another. These systems analyse sentence structure and meaning in the source language and generate equivalent translations in the target language.
Frameworks and Ethical Frameworks
Frameworks are sets of steps that help in solving problems. They give step-by-step guidance for solving problems in an organized way. They also give a structured approach so that relevant factors and considerations are included.
Frameworks also help people communicate and collaborate. They act as a common language. They support the sharing of best practices and bring consistency to problem-solving methods.
Ethics are values or morals that help separate right from wrong. Ethical frameworks combine ethics with step-by-step guidance. They help make sure that choices do not cause unintended harm.
Ethical frameworks also help when moral dilemmas are complex. They allow different ethical principles and perspectives to be considered. By using them, individuals and organizations can make well-informed decisions that match their values and support positive outcomes for all stakeholders involved.
Everyday example: When a student follows fixed steps to complete a school project and checks that nobody is treated unfairly, the student is using the idea of a framework with ethical thinking.
Why AI Needs Ethical Frameworks
AI can produce unwanted outcomes if bias enters the solution. One example is a hiring algorithm that was biased against women applicants. This shows why AI solutions must be checked ethically.
AI is used as a decision-making or decision-influencing tool. Because of this, it should make morally acceptable recommendations. Ethical frameworks help ensure that AI makes morally acceptable choices.
If ethical frameworks are used while building AI solutions, unintended outcomes can be avoided before they happen. This makes ethical thinking important during the creation of AI systems, not only after the solution is already in use.
The "My Goodness" activity, where players decide how to make charitable donations, shows that personal morals, values, and ethics can influence decisions. It also shows that people may have hidden biases in their judgement.
Factors that could influence decisions without being noticed include:
- Identity of the charity recipient.
- Location of the recipient.
- Bias towards relatives.
- Uncovering information available.
Other factors shown as influencing decision-making include:
- Culture
- Value of humans
- Value of non-humans
- Religion
- Intuition & Values
These factors help us understand our own thought processes. Understanding them can help in producing a framework that supports ethically sounder decisions.
Types of Ethical Frameworks
Ethical frameworks for AI are classified into two main types:
- Sector-based Frameworks
- Value-based Frameworks
Sector-based Frameworks are made for specific sectors or industries. In AI, one common sector-based framework is Bioethics. It focuses on ethical considerations in healthcare. It deals with issues such as patient privacy, data security, and the ethical use of AI in medical decision-making. Sector-based ethical frameworks may also apply to finance, education, transportation, agriculture, governance, and law enforcement.
Value-based Frameworks focus on fundamental ethical principles and values that guide decision-making. They reflect different moral philosophies used in ethical reasoning. They assess the moral worth of actions and guide ethical behaviour.
Value-based frameworks are further classified into three categories:
- Rights-based
- Utility-based
- Virtue-based
Rights-based frameworks give priority to protecting human rights and dignity. They value human life over other considerations. They emphasize individual autonomy, dignity, and freedoms. In AI, this means ensuring that AI systems do not violate human rights or discriminate against certain groups.
Utility-based frameworks judge actions by maximizing utility or overall good. They aim to create outcomes that give the greatest benefit and reduce harm. They try to maximize benefit for the greatest number of people. In AI, this may involve weighing the benefits of AI applications against risks to society, such as job displacement or privacy concerns.
Virtue-based frameworks focus on the character and intentions of the people involved in decision-making. They check whether the actions of individuals or organizations match virtues such as honesty, compassion, and integrity. In AI, this can mean checking whether developers, users, and regulators uphold ethical values throughout the AI lifecycle.
These classifications help address ethical concerns in AI development and deployment. They make sure that both sector needs and basic ethical values are considered.
Bioethics and Its Principles
Bioethics is an ethical framework used in healthcare and life sciences. It deals with ethical issues related to health, medicine, and biological sciences. It helps ensure that AI applications in healthcare follow ethical standards and considerations.
The principles of bioethics are:
- Respect for Autonomy.
- Do not harm.
- Ensure maximum benefit for all.
- Give justice.
Respect for Autonomy means that users should be fully aware of decision-making. In the case of an AI algorithm, users should know how it functions.
Do not harm is linked with non-maleficence. Non-maleficence means avoiding harm or negative consequences. It emphasizes reducing harm as much as possible and choosing actions that prevent harm to individuals, communities, or the environment.
Maleficence means intentionally causing harm or wrongdoing. Ethical AI should avoid this.
Ensure maximum benefit for all is linked with beneficence. Beneficence means promoting and maximizing the well-being and welfare of individuals and society. It focuses on actions that create positive outcomes and contribute to the overall good. It also aims to achieve the greatest benefit for all stakeholders involved.
Give justice means that the benefits and burdens of a choice should be distributed in a justified way among people, regardless of their background.
Healthcare AI Case Study Using Bioethics
A company wanted to help hospitals improve patient care. It created an AI algorithm to identify people at high risk. The aim was to help healthcare providers allocate resources effectively and give attention to those most in need.
However, the model could create unintended consequences. Biases or inaccuracies in the data could cause patients to be misclassified or critical cases to be missed. This made the accuracy and reliability of the algorithm very important because flaws in design or training data could affect patient care and outcomes.
The problem was that patients from the Western region of a particular area, who were placed at the same risk level by the algorithm, generally had more severe health conditions than patients from other regions.
The problem happened because:
- The algorithm was trained on healthcare expense data as a measure for health metrics, instead of actual physical illness.
- The algorithm was created in the United States where less money is spent on western region patient healthcare than other ethnic patient healthcare.
The four principles of bioethics can be used to make the healthcare AI solution ethical.
Respect for autonomy requires users to be fully aware of decision-making. The data used to train models and make decisions should be reproducible and accessible to patients. If there are performance concerns, model predictions and data labels should be released.
Do not harm means harm to anyone, human or non-human, must be avoided at all costs. If there is no choice, the path of least harm must be chosen. The solution should promote well-being, minimize harm, and distribute benefits and harms among stakeholders justly. The AI algorithm must be trained on datasets that reduce harm fairly for all groups. In this case, patients from other regions excluding the western part who were less ill would receive more intensive care than patients who actually required help. If implemented, the algorithm would harm patients belonging to the western region by recommending healthcare resource allocation inappropriately.
Maximum benefit means actions should not only avoid harm but also provide as much benefit as possible. The solution should follow clinical practice standards, not only technological ethics standards. It should go beyond nonmaleficence and strive for beneficence. The AI algorithm should not only avoid harming patients from the western region but also benefit them, patients from other regions, and patients of all races. Training data should be unbiased and should reflect healthcare needs and outcomes of patients of all races.
Justice means all benefits and burdens of a choice must be distributed fairly across people, irrespective of background. Solution development needs deep knowledge of social structures that result in issues like racism and sexism, which are types of societal biases. The solution should be aware of social determinants of healthcare and actively work against those structures.
This case study shows that following bioethical principles could help avoid unintended consequences in an AI solution.
About these notes
These notes and questions were written by Syllabax from the official CBSE Class X Artificial Intelligence (417) Facilitator Handbook and the 2026-27 curriculum. This is not an official CBSE publication.