Analysis - Part 2
Basic Concepts & Terminology
[executive summary]
Understanding the following concepts and terminology will greatly demystify key aspects of artificial intelligence. This paves the way to making informed decisions about if and how to invest in educational AI.
While there are no single, universal definitions for these basic terms, the following are working definitions for K-20 educators that I've adopted from a collection of authoritative historical and contemporary sources compiled by UNESCO (2019a) and from an excellent report by Carnegie Mellon professor Robert F. Murphy (Murphy, 2019).
Artificial Intelligence (AI) -"Software algorithms and techniques that allow computers and machines to simulate human perception and decision making processes to successfully complete tasks" (Murphy, 2019, p. 2).
Machine Learning - The function of a computer system designed to gradually be able to perform important tasks by using algorithms to generalize from provided examples; automatically improving with experience.
Deep Learning - A more complex form of machine learning that uses layers of algorithms to be able to generalize without being provided with examples.
Big Data - The practice of working with collections of data that are so large and complex that traditional data management methods are ineffective, requiring newer, more sophisticated data management technologies.
Data Analytics (DA) - The practice of using computer systems to examine raw data so that meaning can be made from them.
Learning Analytics - Data analytics applied to teaching and learning to identify learner habits, predict learner responses, provide timely feedback, support decision-making, simplify realistic assessments and provide personal supervision of learners’ progress.
Adaptive Learning - A computer-based educational system that changes the type of content, the rate of progression through increasingly sophisticated types of content and other content elements in response to student performance on the system.
Examples of established adaptive learning systems include ALEKS, Dreambox Learning, and MATHia.
Apart from understanding the key terms given above, it's important to know that AI in education comes in two broad categories:
While there are no single, universal definitions for these basic terms, the following are working definitions for K-20 educators that I've adopted from a collection of authoritative historical and contemporary sources compiled by UNESCO (2019a) and from an excellent report by Carnegie Mellon professor Robert F. Murphy (Murphy, 2019).
Artificial Intelligence (AI) -"Software algorithms and techniques that allow computers and machines to simulate human perception and decision making processes to successfully complete tasks" (Murphy, 2019, p. 2).
Machine Learning - The function of a computer system designed to gradually be able to perform important tasks by using algorithms to generalize from provided examples; automatically improving with experience.
Deep Learning - A more complex form of machine learning that uses layers of algorithms to be able to generalize without being provided with examples.
Big Data - The practice of working with collections of data that are so large and complex that traditional data management methods are ineffective, requiring newer, more sophisticated data management technologies.
Data Analytics (DA) - The practice of using computer systems to examine raw data so that meaning can be made from them.
Learning Analytics - Data analytics applied to teaching and learning to identify learner habits, predict learner responses, provide timely feedback, support decision-making, simplify realistic assessments and provide personal supervision of learners’ progress.
Adaptive Learning - A computer-based educational system that changes the type of content, the rate of progression through increasingly sophisticated types of content and other content elements in response to student performance on the system.
Examples of established adaptive learning systems include ALEKS, Dreambox Learning, and MATHia.
Apart from understanding the key terms given above, it's important to know that AI in education comes in two broad categories:
AI Category |
Features |
Examples |
Rule-Based / Knowledge-Driven Systems |
Make decisions based on expert processing rules programmed by humans |
Simpler adaptive learning systems, intelligent tutoring systems |
Machine Learning / Data-Driven Systems |
Make predictions using auto-generated rules and based on analysis of large, rich data sets |
More complex adaptive learning systems, automated essay scoring, early warning systems for poor academic performance and late graduation or dropout |
[closer look]
Understanding the following concepts and terminology will greatly demystify key aspects of artificial intelligence. This paves the way to making informed decisions about if and how to invest in educational AI.
While there are no single, universal definitions for these basic terms, the following are working definitions for K-20 educators that I've adopted from a collection of authoritative historical and contemporary sources compiled by UNESCO (2019a) and from an article by Carnegie Mellon professor Robert F. Murphy.
Artificial Intelligence (AI) -"Software algorithms and techniques that allow computers and machines to simulate human perception and decision making processes to successfully complete tasks" (Murphy, 2019, p. 2).
Narrow AI / Weak AI - Software that performs a single, specific function (e.g. Google Assistant responding to a question about today's weather; a driverless car recognizing a stop sign as different from a yield sign)
Strong AI / General Intelligence AI - Theoretical (non-existent) software that approaches the cognitive reasoning abilities of humans.
Machine Learning - The function of a computer system designed to gradually be able to perform important tasks by using algorithms to generalize from provided examples; automatically improving with experience.
When a machine "learns" it is using statistical algorithms to create a prediction model by processing large amounts of rich, varied data about the subject its learning.
Deep Learning - A more complex form of machine learning that uses layers of algorithms to be able to generalize without being provided with examples.
Big Data - The practice of working with collections of data that are so large and complex that traditional data management methods are ineffective, requiring newer, more sophisticated data management technologies.
There are "four Vs" that typically need to be considered with big data:
Volume: how much data are there?
Variety: how diverse are the data?
Velocity: how quickly are new data generated?
Veracity: how accurate are the data?
Data Analytics (DA) - The practice of using computer systems to examine raw data so that meaning can be made from them.
Learning Analytics - Data analytics applied to teaching and learning to identify learner habits, predict learner responses, provide timely feedback, support decision-making, simplify realistic assessments and provide personal supervision of learners’ progress.
Adaptive Learning - A computer-based educational system that changes the type of content, the rate of progression through increasingly sophisticated types of content and other content elements in response to student performance on the system.
Examples of established adaptive learning systems include ALEKS, Dreambox Learning, and MATHia.
Apart from the understanding the key terms given above, it's important to know that AI in education comes in two broad categories:
While there are no single, universal definitions for these basic terms, the following are working definitions for K-20 educators that I've adopted from a collection of authoritative historical and contemporary sources compiled by UNESCO (2019a) and from an article by Carnegie Mellon professor Robert F. Murphy.
Artificial Intelligence (AI) -"Software algorithms and techniques that allow computers and machines to simulate human perception and decision making processes to successfully complete tasks" (Murphy, 2019, p. 2).
Narrow AI / Weak AI - Software that performs a single, specific function (e.g. Google Assistant responding to a question about today's weather; a driverless car recognizing a stop sign as different from a yield sign)
Strong AI / General Intelligence AI - Theoretical (non-existent) software that approaches the cognitive reasoning abilities of humans.
Machine Learning - The function of a computer system designed to gradually be able to perform important tasks by using algorithms to generalize from provided examples; automatically improving with experience.
When a machine "learns" it is using statistical algorithms to create a prediction model by processing large amounts of rich, varied data about the subject its learning.
Deep Learning - A more complex form of machine learning that uses layers of algorithms to be able to generalize without being provided with examples.
Big Data - The practice of working with collections of data that are so large and complex that traditional data management methods are ineffective, requiring newer, more sophisticated data management technologies.
There are "four Vs" that typically need to be considered with big data:
Volume: how much data are there?
Variety: how diverse are the data?
Velocity: how quickly are new data generated?
Veracity: how accurate are the data?
Data Analytics (DA) - The practice of using computer systems to examine raw data so that meaning can be made from them.
Learning Analytics - Data analytics applied to teaching and learning to identify learner habits, predict learner responses, provide timely feedback, support decision-making, simplify realistic assessments and provide personal supervision of learners’ progress.
Adaptive Learning - A computer-based educational system that changes the type of content, the rate of progression through increasingly sophisticated types of content and other content elements in response to student performance on the system.
Examples of established adaptive learning systems include ALEKS, Dreambox Learning, and MATHia.
Apart from the understanding the key terms given above, it's important to know that AI in education comes in two broad categories:
AI Category |
Features |
Examples |
Rule-Based / Knowledge-Driven Systems |
Make decisions based on expert processing rules programmed by humans |
Simpler adaptive learning systems, intelligent tutoring systems |
Machine Learning / Data-Driven Systems |
Make predictions using auto-generated rules and based on analysis of large, rich data sets |
More complex adaptive learning systems, automated essay scoring, early warning systems for poor academic performance and late graduation or dropout |
Expert Advice
[executive summary]
When it comes to educational AI, I consider "experts" to be people who have done original scientific research on the topic, have thoroughly studied the topic academically and/or who have been involved in the development of educational AI systems.
The "advice" presented here is a summary of expert opinions compiled from recently published works listed in the references section of this website.
What's Important to Know
Focus Your AI Investment
It may be helpful to decide early on, which of three general categories of AI you want to invest in:
Practical Educational AI
Scrutinize AI Providers
The "advice" presented here is a summary of expert opinions compiled from recently published works listed in the references section of this website.
What's Important to Know
- Rule-Based / Knowledge-Based applications (simpler adaptive learning systems, intelligent tutoring systems)
- have been proven to improve test scores across grade levels and subjects areas, compared to traditional teacher-led instruction
- have shown results similar to one-on-one tutoring and small-group instruction
- mostly apply to math, literacy, physical sciences and computer science
- are best for learning content that is suited to rules-based approaches: facts, methods, operations, and procedural skills
- are best used for independent, remedial or enrichment activities, not in-class activities
- Machine Learning-based / Data-Driven applications (more complex adaptive learning systems, automated essay scoring, early warning systems)
- have proven significantly better at predicting late graduation and dropout rates, compared to rules-based approaches
- have potential to provide superior adaptive learning systems, if designed very carefully
- are not yet as mature in education as rules-based AI products
Focus Your AI Investment
It may be helpful to decide early on, which of three general categories of AI you want to invest in:
- AI management tools focused on administration (data analytic and predictive tools for enrollment, grade- and division-level academic performance, graduation and dropout rates, socioeconomic factors, etc.)
- AI learning tools focused on students (adaptive learning systems, intelligent tutoring systems, etc.)
- AI teaching tools focused on teachers (lesson preparation tools, automated student administration, evaluation and feedback systems, automated essay assessment, etc.)
Practical Educational AI
- For administrators, AI offers data-based reports and "dashboards" that can assist with managing and administering a district or school more efficiently, developing feasible and cost-effective plans, formulating responsive policies, and monitoring and evaluating educational outcomes
- For students, AI is well suited to adaptive learning; understanding what students know and don't know
- AI is not well suited (yet) to personalized learning, which involves understanding what students want to know and how they learn best
- For faculty, AI currently has the least potential to help in areas where teachers are directly engaging with students
- it's best applied as tools that can save teacher time by automating or streamlining administrative and organizational tasks, and that assist teachers to provide students with more meaningful feedback and individualized attention
Scrutinize AI Providers
- Assess whether the AI product/service you're considering is using a rules-based approach or a machine learning-based approach
- Review the strengths and weaknesses of the approach (see information above)
- Challenge the AI provider with questions you have based on your assessment
- Prioritize providers that have a history of working with non-profit researchers
- Only invest in companies that base their claims in scientific/academic research that is directly relevant to their product's features, functions and intended uses
[closer look]
When it comes to educational AI, I consider "experts" to be people who have done original scientific research on the topic, have thoroughly studied the topic academically and/or who have been involved in the development of educational AI systems.
The "advice" presented here is a summary of expert opinions compiled from recently published works listed in the references section of this website.
What's Important to Know
Focus Your AI Investment
It may be helpful to decide early on, which of three general categories of AI you want to invest in:
Practical Educational AI
Scrutinize AI Providers
The "advice" presented here is a summary of expert opinions compiled from recently published works listed in the references section of this website.
What's Important to Know
- Rule-Based / Knowledge-Based applications (simpler adaptive learning systems, intelligent tutoring systems)
- have been proven to improve test scores across grade levels and subjects areas, compared to traditional teacher-led instruction
- have shown results similar to one-on-one tutoring and small-group instruction
- mostly apply to math, literacy, physical sciences and computer science
- are best for learning content that is suited to rules-based approaches: facts, methods, operations, and procedural skills
- are best used for independent, remedial or enrichment activities, not in-class activities
- are not good for learning complex, difficult to assess, higher-order skills: critical thinking, effective communication, explanation, argumentation, collaboration, self-management, social awareness, etc.
- require that time be allocated for teachers to regularly review reports on student progress and performance
- Machine Learning-based / Data-Driven applications (more complex adaptive learning systems, automated essay scoring, early warning systems)
- have proven significantly better at predicting late graduation and dropout rates, compared to rules-based approaches
- have potential to provide superior adaptive learning systems, if designed very carefully
- are not yet as mature in education as rules-based AI products
- large, high-quality data sets needed to train the systems are difficult to obtain (e.g. training sets should represent the diversity of the application’s target population)
- statistical models will encode any biases that might be embedded in the training data
- lack of transparency about a company's AI models (to protect intellectual property) can easily damage public trust in a product
Focus Your AI Investment
It may be helpful to decide early on, which of three general categories of AI you want to invest in:
- AI management tools focused on administration (data analytic and predictive tools for enrollment, grade- and division-level academic performance, graduation and dropout rates, socioeconomic factors, etc.)
- AI learning tools focused on students (adaptive learning systems, intelligent tutoring systems, etc.)
- AI teaching tools focused on teachers (lesson preparation tools, automated student administration, evaluation and feedback systems, automated essay assessment, etc.)
Practical Educational AI
- For administrators, AI offers data-based reports and "dashboards" that can assist with managing and administering a district or school more efficiently, developing feasible and cost-effective plans, formulating responsive policies, and monitoring and evaluating educational outcomes
- For students, AI is well suited to adaptive learning; understanding what students know and don't know
- AI is not well suited (yet) to personalized learning, which involves understanding what students want to know and how they learn best
- For faculty, AI currently has the least potential to help in areas where teachers are directly engaging with students
- it's best applied as tools that can save teacher time by automating or streamlining administrative and organizational tasks, and that assist teachers to provide students with more meaningful feedback and individualized attention
Scrutinize AI Providers
- Assess whether the AI product/service you're considering is using a rules-based approach or a machine learning-based approach
- Review the strengths and weaknesses of the approach (see information above)
- Challenge the AI provider with questions you have based on your assessment
- Prioritize providers that have a history of working with non-profit researchers
- Only invest in companies that base their claims in scientific/academic research that is directly relevant to their product's features, functions and intended uses.
Further Reading
Magazine Articles
AI in schools — here’s what we need to consider (The Conversation)
China has started a grand experiment in AI education. It could reshape how the world learns. (MIT Technology Review)
Examples of Artificial Intelligence in Education (Emerj)
Flawed Algorithms Are Grading Millions of Students’ Essays (MotherBoard)
How Is AI Used In Education — Real World Examples Of Today And A Peek Into The Future (Forbes)
‘It’s an educational revolution’: how AI is transforming university life (The Guardian)
Reports / White Papers
From Good Intentions to Real Outcomes: Equity by Design in Learning Technologies (Connected Learning Alliance)
Human Centered AI Index Report (Stanford University)
AI in schools — here’s what we need to consider (The Conversation)
China has started a grand experiment in AI education. It could reshape how the world learns. (MIT Technology Review)
Examples of Artificial Intelligence in Education (Emerj)
Flawed Algorithms Are Grading Millions of Students’ Essays (MotherBoard)
How Is AI Used In Education — Real World Examples Of Today And A Peek Into The Future (Forbes)
‘It’s an educational revolution’: how AI is transforming university life (The Guardian)
Reports / White Papers
From Good Intentions to Real Outcomes: Equity by Design in Learning Technologies (Connected Learning Alliance)
Human Centered AI Index Report (Stanford University)
Feedback / Suggestions
I welcome your thoughts, questions, critiques and ideas for improving this OER.
I welcome your thoughts, questions, critiques and ideas for improving this OER.