Decision-Making in Machines/Computers

1. Autonomous system
2. Automated system
3. Automated system

1. Human decision-making
2. Machine decision-making
3. Human decision-making

1. Selecting a dress for a wedding — Subjective

2. Calculating the factorial of a given number — Not subjective (objective)

3. Selecting a restaurant for dinner — Subjective

4. Selecting the location for opening a shop — Subjective

5. Choosing a new business — Subjective

6. Finding the best doctor for treatment — Subjective

True

Reason — Automated systems mainly use deterministic computing because they do not learn from their environment; they work on predefined rules. Probabilistic computing is more suitable for autonomous systems. Since they learn from the environment, there is some uncertainty and randomness.

False

Reason — Human decision-making is often subjective, which means that it is affected by individual prejudices, feelings, and experiences. Machine decision-making is mainly objective, which means that it is unaffected by biases or feelings. Hence, human decision-making is not always subjective and machine decision-making is not always objective.

False

Reason — Humans classify objects based on their different attributes, like colour, shape, texture, appearance, application, etc. Humans learn all these things through experience. Machines identify objects by using machine learning algorithms and related datasets for training and testing purposes. Hence, the classification process of humans and machines is not identical.

True

Reason — Machine learning is a subset of Artificial Intelligence (AI) that enables software programs to become more accurate predictors of outcomes without being explicitly programmed to do so.

True

Reason — Data works as the input of a machine learning model, and information is the processed form of data that helps in understanding patterns. This enables the machine learning model to learn and make decisions and predictions.

True

Reason — Python is the most commonly used programming language for machine learning. The following features of Python make it suitable for machine learning: Open source, huge libraries of machine learning tasks, easy syntax, and can run on cloud platform.

False

Reason — Programming plays a significant role in subjective decision-making by providing a structured approach to analyse, model, and implement decisions influenced by personal opinions, preferences, and judgement.

True

Reason — Sorting fruits involves subjective factors such as taste, colour, shape, size, and personal preference, and machine learning helps in handling these factors during the decision-making process.

True

Reason — Machine learning is used in a variety of applications such as healthcare for disease prediction, finance for analysis and decision-making, and transportation for systems like autonomous vehicles and traffic management.

True

Reason — Machine learning is a rapidly evolving field in which new techniques, models, and applications are continuously being developed to improve decision-making and performance.

Autonomous systems are adaptable.

Reason — Autonomous systems can learn from their environment and have decision-making capability. This makes them more flexible and adaptable, unlike automated systems which work only on predefined rules.

A robot that assembles cars on an assembly line

Reason — Autonomous systems have decision-making capability and can work with a certain level of independence by learning from their environment. Robots are listed as examples of autonomous systems, whereas devices like refrigerators and washing machines are automated systems that work on predefined rules.

Adaptive and subjective

Reason — Human decision-making is influenced by emotions, experiences, and prejudices, which makes it subjective, and humans can learn from past experiences and change their decisions according to different situations, which makes it adaptive.

Data is used to train models and make predictions.

Reason — In machine learning, data works as the input for training algorithms. By analysing data, the model learns patterns and later uses this learning to make predictions and decisions.

All of the above

Reason — Subjective decisions are influenced by personal preferences, opinions, experiences, and judgement. Choosing fruits, deciding to use an air conditioner, and investing in stocks all depend on individual perceptions and circumstances.

To identify objects based on their characteristics

Reason — Object classification means identifying or recognising objects by using their features such as colour, shape, texture, appearance, and other attributes.

All of the above

Reason — Programming is required to implement algorithms, and algorithms provide the step-by-step process that enables machines to learn from data and perform machine learning tasks. Hence, programming and algorithms are both essential for machine learning.

Object detection

Reason — Machine learning can be used to detect objects in an image or video. Identifying a car from an image is an example of detecting an object using its visual features.

Input of a machine learning algorithm

Reason — Data works as the input of a machine learning model, which is used for learning patterns and making predictions and decisions.

All of the above

Reason — The machine learning process includes gathering data, selecting and implementing algorithms, training the model, and evaluating or analysing the results before making predictions.

1. Automated systems are characteristically deterministic, while autonomous systems are deterministic or probabilistic.
2. Sorting fruits is an illustration of using machine learning to make subjective decisions.
3. Object classification is the task of assigning a label to an object.
4. Machine learning trusts on analysing data to make decisions and predictions.
5. In decision-making, humans frequently consider subjective factors.

Subjective decision-making is influenced by personal feelings, emotions, experiences, beliefs, and preferences of an individual. In contrast, objective decision-making is based on facts, data, logic, and predefined rules, and is not affected by personal biases or emotions.

Automated decision-making processes are based on predefined rules and programs and do not learn from the environment. They are deterministic and objective in nature. Human decision-making, on the other hand, is often subjective, influenced by emotions, experiences, and personal prejudices, and humans can adapt their decisions based on different situations.

Autonomous system implementations present difficulties related to safety, privacy, accountability, and transparency. It is challenging to ensure that autonomous systems do not harm humans, protect personal data, clearly define responsibility when mistakes occur, and make their decision-making process understandable to humans.

An automated system works on predefined rules and programs written by humans. It cannot learn from its environment and does not have decision-making capability.

An autonomous system, on the other hand, can learn from its environment, has decision-making capability, and can adapt its actions according to different situations.

Algorithms and programming are both very important for machine learning. Programming is needed to write code that applies machine learning methods, and algorithms provide the step-by-step rules that help a machine learn patterns from data and make predictions or decisions. Hence, both are essential for machine learning.

The stages in machine learning are as follows:

1. Data collection — This is a very important step. To prepare a machine learning software, we need to collect related data.
2. Data preparation — The collected data may not be in a form to be entered directly into the algorithm. To make it compatible with the algorithm, we need to clean and transform it.
3. Selection of algorithm — There are many machine learning algorithms. We need to choose an appropriate machine learning algorithm as per our data and requirements.
4. Training — In this step, data is provided to the selected algorithm for training. By analysing and observing data, the algorithm learns its important features.
5. Evaluation — To check whether the model has learnt the features or not, we need to test it. At the time of testing, we check its accuracy and evaluate the performance of the algorithm.
6. Predict or deploy — If the testing performance of the model is good, then we can use it to make predictions.

The potential drawbacks or difficulties of using machine learning in decision-making are:

1. Machines are not able to learn from experience in the same way that humans do. This means that machines may not be able to make the same kind of creative and flexible decisions that humans can.
2. It is difficult to quantify human emotions and values through algorithms or programming. As a result, machine learning systems may face difficulty in making subjective decisions that involve emotions, beliefs, and personal values.

There are three main types of machine learning:

1. Supervised learning: The machine learning program is trained on a set of data that has already been labelled. By matching the names to the data, the algorithm learns to find trends in the data.
2. Unsupervised learning: The machine learning program is trained on a set of data that has not been labelled. The program learns to find patterns in the data by looking for groups of patterns that occur at the same time.
3. Reinforcement learning: In this method, an agent interacts with its surroundings and learns from trial and error. The agent takes actions, receives feedback, and adjusts its behaviour to maximise cumulative rewards.

The difficulties in creating autonomous systems are:

1. Safety: The system must work safely in all situations and should not harm people or property.
2. Accuracy and reliability: It should give correct results consistently, even with new or unexpected inputs.
3. Sensing and understanding the environment: Using sensors/cameras to detect objects, distance, and obstacles correctly is difficult, especially in poor light, rain, dust, etc.
4. Decision making in real time: The system must take quick and correct decisions without delay.
5. Handling unexpected situations: Real life has surprises (sudden obstacles, unusual behaviour of people/vehicles), which are hard to predict.
6. Privacy: Protecting people’s personal data collected by the system (images, location, etc.) is challenging.
7. Security: Autonomous systems can be hacked, so strong protection is required.
8. Accountability: It is difficult to decide who is responsible if the system makes a mistake—developer, company, or user.
9. Transparency: The system’s decisions should be explainable so humans can understand and trust them.
10. Cost and maintenance: Sensors, software updates, and repairs can be expensive and need regular maintenance.

The task of sorting fruits may be considered as subjective decision-making when we are sorting fruits for personal use. In this case, the following may be the decision-making factors:

• Taste: The taste of the fruits may be the deciding factor. For example, different people like different varieties of mangoes.
• Shape or size: At the time of sorting fruits, we may be conscious of the shape of the fruit. Size is also an important factor at the time of sorting. We may prefer to collect fruits of the same size.
• Colour: Colour is also a deciding factor. For example, we only like red apples.
• Price: It would be a quantitative criterion for decision-making.

Sorting fruits involves a combination of subjective factors like personal preference as well as objective factors like quality standards and price.

Machine learning can improve subjective decision-making by making it more data-driven and consistent. It helps in the following ways:

1. Finds patterns from past data: ML studies previous decisions and outcomes to learn what usually leads to better results.
2. Converts opinions into numbers: Feedback like ratings, reviews, and choices can be changed into scores, so they can be compared more objectively.
3. Reduces human bias: The same rules are applied to everyone, which can reduce favouritism and random judgement.
4. Supports better comparisons: ML can compare many factors at the same time and suggest the best option.
5. Gives predictions and recommendations: It can predict likely outcomes and recommend actions, helping decision makers choose wisely.

Thus, machine learning makes subjective decisions more fair, accurate, and based on evidence.

To create a machine learning model that predicts whether a student will be successful in an exam, we need a dataset containing information about students from the past. This data helps the model learn patterns and make predictions.

The required data can include:

1. Academic Performance Data

• Marks/grades in previous tests and exams
• Subjec-wise scores and overall percentage

2. Study and Practice Data

• Hours spent studying regularly
• Homework and assignment completion
• Number of practice papers/tests attempted

3. Attendance Data

• Attendance percentage
• Regularity in attending classes

4. Classwork/Behaviour Indicators

• Class participation
• Teacher feedback on performance and effort

5. Final Result (Target Data)

• The past final exam result of the student (e.g., Pass/Fail or Successful/Not successful)
• This is the output label that the model learns to predict.