3.2A Characteristics of an Algorithm

Unambiguous: Each step in an algorithm must be clear and have only one interpretation.

Examples:

• Face recognition algorithms on smartphones that follow precise mathematical procedures to identify facial features.

• QR code scanning algorithms that have explicit steps for detecting and decoding patterns.

• Secure password hashing algorithms that follow exact steps to convert passwords into encrypted strings.

Finite: An algorithm must terminate after a finite number of steps.

Examples:

• Netflix's recommendation algorithm that processes user data through a defined number of steps.

• Credit card validation algorithms that check digits with a limited number of operations.

• Digital signature verification algorithms that confirm authenticity in a fixed number of steps.

Well-defined inputs and outputs: The algorithm must have clearly specified inputs and produce expected outputs.

Examples:

• Weather prediction algorithms that take meteorological data as input and output forecasts.

• Currency conversion apps that take amount and currency types as inputs and output converted values.

• Language translation algorithms that take text in one language as input and output text in another.

Feasible: The algorithm must be executable with available resources.

Examples:

• Mobile banking apps using encryption algorithms designed for smartphone processing capabilities.

• Video compression algorithms that reduce file size while maintaining acceptable quality.

• Web browser rendering algorithms optimized to display pages quickly on various devices.

3.2B Components of an Algorithm

Instructions: The specific steps the algorithm follows.

Examples:

• GPS navigation instructions that guide drivers through specific turns and routes.

• Digital photo editing algorithms with step-by-step processes for adjusting brightness, contrast, etc.

• E-commerce checkout processes that follow specific steps from cart to payment confirmation.

Variables: Named storage locations for data that can change during algorithm execution.

Examples:

• Social media algorithms storing user engagement metrics as variables.

• Weather apps tracking changing temperature, humidity, and pressure variables.

• Investment apps storing portfolio value variables that update with market fluctuations.

Conditionals: Decision points where the algorithm chooses different paths based on conditions.

Examples:

• Email spam filters using conditionals to classify messages based on content analysis.

• Adaptive learning platforms that adjust difficulty based on student performance.

• Banking security systems that flag transactions if they match suspicious patterns.

Loops: Repetitive execution of instructions until a condition is met.

Examples:

• Fitness tracking apps repeatedly counting steps until movement stops.

• Video streaming buffering algorithms that loop to load content until sufficient data is cached.

• Web crawlers that repeatedly follow links until they've indexed all pages on a site.

3.2C Ways of Representing Algorithms

Natural language: Algorithm steps described in everyday language.

Examples:

• Recipe apps presenting cooking instructions in plain language steps.

• Customer support chatbots following natural language decision trees.

• User manuals describing software functionality in plain language.

Flow chart: Graphical representation showing steps as shapes connected by arrows.

Examples:

• Visual programming interfaces like Scratch representing code as connected blocks.

• Business process modeling tools showing workflow as connected shapes.

• Troubleshooting guides with flowcharts for resolving technical issues.

Code: Algorithm written in a formal programming language.

Examples:

• YouTube's video recommendation system written in Python.

• Instagram's image processing filters implemented in Swift for iOS.

• LinkedIn's job matching algorithms written in Java.

Programming languages: Formal languages used to implement algorithms.

Examples:

• JavaScript used to create interactive web applications.

• Python used for data analysis and machine learning applications.

• SQL used for database query algorithms.

3.2D Uses of Algorithms

Sorting, searching, filtering, prioritizing, classifying, associating, counting

Sorting Examples:

• Spotify's algorithms organizing music playlists by various criteria.

• Email clients sorting messages by date, sender, or priority.

• E-commerce sites sorting products by price, popularity, or relevance.

Searching Examples:

• Google's search algorithm finding relevant web pages based on keywords.

• Shazam's audio fingerprinting algorithm identifying songs from snippets.

• Netflix's search algorithm finding content based on titles, actors, or genres.

Filtering Examples:

• Instagram's content filters removing inappropriate material.

• Noise-cancellation algorithms filtering out background sounds.

• Ad blockers filtering unwanted advertisements on websites.

Prioritizing Examples:

• Email inbox algorithms highlighting important messages.

• News feed algorithms prioritizing content based on user interests.

• Notification systems prioritizing alerts based on urgency.

Classifying Examples:

• Amazon's product categorization algorithms.

• Gmail's inbox categorization (Primary, Social, Promotions).

• YouTube's video content classification system.

Associating Examples:

• LinkedIn's "People You May Know" algorithm connecting related professionals.

• Spotify's "Discover Weekly" associating music with listener preferences.

• Amazon's "Frequently Bought Together" product associations.

Counting Examples:

• Website analytics tools counting visitor interactions.

• Social media engagement counters for likes, shares, and comments.

• Step counters in fitness apps.

Programming, software development and implementation

Examples:

• GitHub Copilot suggesting code completions to programmers.

• Integrated Development Environments (IDEs) with code analysis algorithms. An Integrated Development Environment (IDE) is a comprehensive software application that provides programmers with a complete set of tools for software development in one unified interface. IDEs combine multiple development tools into a single framework to streamline the coding process.

• Automated testing frameworks that verify software functionality.

• Compiler optimization algorithms that improve code efficiency.

Machine learning, neural networks and creation of other algorithms

Machine Learning Examples:

• Tesla's self-driving car algorithms improving through experience.

• Voice assistants like Siri and Alexa improving speech recognition over time.

• Fraud detection systems that learn new patterns of suspicious activity.

Neural Networks Examples:

• DeepL translation service using neural networks for accurate translations. DeepL is an AI-based translation service that was launched in 2017 by the German company DeepL GmbH (formerly Linguee GmbH). It uses neural networks and deep learning technologies to provide high-quality translations between multiple languages.

• Image recognition systems like Google Lens identifying objects in photos.

• Medical diagnosis systems analyzing medical images for disease patterns.

Algorithm Creation Examples:

• AutoML platforms automatically generating machine learning algorithms. AutoML (Automated Machine Learning) platforms are software systems designed to automate the process of creating, optimizing, and deploying machine learning models with minimal human intervention. These platforms enable users with varying levels of data science expertise to build effective AI models.

• Reinforcement learning systems that develop their own strategies.

3.2E Algorithmic Dilemmas

Algorithmic bias and fairness

Examples:

• Facial recognition systems performing less accurately for certain demographic groups.

• Resume screening algorithms potentially favoring certain language patterns.

• Criminal risk assessment algorithms showing bias in predictions.

• Healthcare algorithms allocating resources based on historical data that may contain biases.

Algorithmic accountability and transparency, black box algorithms

Accountability Examples:

• Credit scoring algorithms determining loan eligibility without clear explanations.

• College admissions algorithms making decisions with limited oversight.

• Healthcare resource allocation algorithms during crisis situations.

Transparency Examples:

• GDPR's "right to explanation" for automated decisions affecting EU citizens.

• Open source algorithms allowing public scrutiny of code.

• Algorithm impact assessments for government-used AI systems.

Black Box Examples:

• TikTok's content recommendation algorithm with undisclosed inner workings.

• Proprietary trading algorithms used in financial markets.

• Deep learning medical diagnosis systems whose reasoning cannot be easily explained.

Erosion and/or loss of human judgment

Examples:

• Automated hiring systems screening candidates without human review.

• Predictive policing algorithms directing law enforcement resources.

• Content moderation algorithms making censorship decisions at scale.

• Automated customer service systems replacing human representatives.

• Medical diagnosis algorithms potentially overriding clinical judgment.

IB DP Digital Society - Section 3.2 Algorithms Practice Questions

Define/State Questions

1. Define the term "algorithm" as used in digital society.

2. State four essential characteristics of an effective algorithm.

3. Define the term "black box algorithm" and state one potential concern associated with it.

4. State three ways algorithms can be represented.

5. Define "algorithmic bias" and state one digital technology where this might occur.

6. State four different components that are typically found in algorithms.

7. Define what "unambiguous" means as a characteristic of algorithms.

8. State three common uses of sorting algorithms in digital society.

9. Define what is meant by "algorithmic accountability."

10. State two ways that machine learning algorithms differ from traditional algorithms.

Identify Questions

11. Identify three types of filtering algorithms used in digital society applications.

12. Identify four examples of how algorithms are used for classification in social media platforms.

13. Identify two ways that algorithmic decision-making might lead to an erosion of human judgment.

14. Identify three components of a neural network algorithm.

15. Identify four situations where algorithmic transparency would be particularly important in digital society.

Outline Questions

16. Outline three ways in which algorithms are used in the creation of other algorithms.

17. Outline the key differences between representing algorithms in natural language versus in code.

18. Outline two potential consequences of using black box algorithms in criminal justice systems.

19. Outline three relationships between algorithmic bias and algorithmic fairness.

20. Outline how conditionals and loops function as components of algorithms.

Describe Questions

21. Describe how the "finite" characteristic of algorithms applies to recommendation systems.

22. Describe two ways that algorithms are represented in visual programming environments.

23. Describe how algorithms are used in the process of prioritizing content on social media platforms.

24. Describe three examples of how automated decision-making algorithms might impact employment practices.

25. Describe how the concept of "well-defined inputs and outputs" applies to facial recognition algorithms.

Explain Questions

26. Explain how algorithmic bias might emerge in facial recognition technologies.

27. Explain two ways that programming languages influence algorithm implementation.

28. Explain three potential impacts of using algorithms for content moderation on social media platforms.

29. Explain how machine learning algorithms differ from traditional algorithms in terms of human intervention.

30. Explain two ways that algorithmic transparency could be improved in credit scoring systems.

Compare Questions

31. Compare two different ways of representing algorithms, focusing on their suitability for different audiences.

32. Compare how sorting algorithms and filtering algorithms are used in digital society applications.

33. Compare the role of human judgment in traditional decision-making versus algorithmic decision-making.

34. Compare two different approaches to addressing algorithmic bias in technology applications.

35. Compare the benefits and limitations of using AutoML platforms versus traditional algorithm development.

Suggest Questions

36. Suggest two ways that algorithmic accountability could be improved in autonomous vehicle technologies.

37. Suggest three potential solutions to the problem of algorithmic bias in hiring algorithms.

38. Suggest two ways that loops in algorithms can be optimized for efficiency.

39. Suggest three criteria that could be used to evaluate the transparency of algorithms used in public services.

40. Suggest how developers might balance efficiency and fairness when designing recommendation algorithms.