Introducing AI in Academia (2 of 4): Using AI to Enhance Learning

Some Current Applications of AI

This is the second in a series of blog posts written by the Artificial Intelligence Faculty Learning Community Group (AI FLC) at Regis College. Although the posts are primarily intended for the Regis community, the conversation is open to everyone. This post was written by Allen Karsina.

A "Deep Dive" into NotebookLM

Welcome back! In our previous post, we identified five key topics that deserve thoughtful discussion:

1. What is AI?

2. How is AI currently being used in education?

3. What are the benefits and opportunities of integrating AI into education?

4. What are the challenges and risks of integrating AI into education?

5. What are the best practices for integrating AI into education?

We explored the first question in our last post, so now we are ready to take a deep dive into the second question - how is AI currently being used in education?

A Quick Look at AI in Higher Education

As we consider how to integrate AI into our schools and classrooms, it is essential to recognize that AI is already an integral part of our daily lives and educational systems. Narrow applications of AI, such as spell check, grammar correction, slideshow captioning, and countless online tools and services, have been supporting students and educators for years.

Recent advances in AI have further expanded its presence. Familiar products are now receiving significant AI enhancements. For example, Zoom offers AI-powered meeting summaries and detailed feedback for recorded presentations, Adobe PDF Reader includes an AI Assistant, Grammarly has integrated advanced writing tools; and Microsoft is integrating Copilot into all of its Microsoft 365 applications. Even a simple Google search now incorporates AI technology. These examples highlight just a fraction of the ways AI is becoming ubiquitous.

I highlight this for two key reasons. First, it is becoming increasingly impractical, if not impossible, to enforce a “no AI” policy in educational settings, given AI’s pervasive integration into common tools and platforms. Second, as AI continues to influence more aspects of our lives, it is crucial for educators to incorporate lessons on the ethical and effective use of AI into the curriculum. Preparing students to navigate this rapidly evolving landscape responsibly and skillfully should be a fundamental part of modern education.

NotebookLM

Rather than trying to exhaustively list all the ways AI is currently being used in education, I will do a "deep dive" into one application from Google Labs called NotebookLM. I chose this application because it contains a number of features that many other applications have, but also at least one that is unique (to my knowledge) and may surprise you. It is also available to both educators and students in free and paid versions, and is easy to access and to use, and therefore some students or educators may already be using it.

Disclaimer - although I chose NotebookLM to highlight in this post, there are many similar services and more on the way. Even NotebookLM's audio overview turns out to be not all that unique. If you are an educator evaluating tools that you might use in your classroom, I recommend prioritizing tools that all of your students can access and that are consistent with your institution's use of technology. For example, students and faculty at Regis have free access to the Enterprise version of Copilot through their Regis accounts. NotebookLM is not supported by Regis College.

What is NotebookLM?

NotebookLM is an online tool that allows you to create notebooks (i.e., folders) with one or more "source". Sources include files, links, and even copied and pasted text that the user uploads to the notebook. Once a source is uploaded and selected, the user can "interact" with the source by clicking on buttons to generate a study guide with answers, a timeline, a briefing document, or a FAQ. Additionally, the user can click on one of three generated questions or enter their own question (or prompt) into a chat bar.

The feature that really shines in NotebookLM, though, is its Audio Overview. With the press of a single button, you can create a podcast-like discussion with two AI hosts. As of December 13th, you can even click a join button and join in the "discussion". For anyone who likes to listen to podcasts or audiobooks, this feature is quite amazing.

[For Google Lab's description of NotebookLM, see here.]

Some Examples Using NotebookLM

To demonstrate a few of the uses of NotebookLM, I uploaded 10 articles and 2 e-books related to Artificial Intelligence in education to a folder in NotebookLM (each folder can have up to 50 sources in the free version).

The following summary was automatically generated:

These sources explore the implications of generative artificial intelligence (GenAI), particularly ChatGPT, in higher education. They examine both the opportunities GenAI offers for enhancing teaching and learning, such as personalized feedback and automated assessment, and the significant ethical challenges it presents, including bias, plagiarism, and data privacy concerns. The authors discuss the need for AI literacy among educators and students, as well as the development of effective AI policies within educational institutions. Several sources also explore the technical aspects of GenAI models, their development, and their limitations. Finally, some sources analyze existing ethical frameworks and propose solutions for responsible AI integration in education and beyond.

Audio Overview

I then clicked on the "generate" button to create an audio overview (see below).

Artificial Intelligence in Education - NotebookLM Audio Overview

0:00

I have created and listened to a lot of audio overviews ("deep dives") across several different topics. They aren't perfect and suffer from the same flaws and potential challenges as text generated by Generative AI. They definitely don't replace reading the source materials. You can customize them to a degree but for the most part you get what you get. Even so, this is an amazing tool with many potential uses.

NotebookLM's "Chat" Feature

Like Adobe PDF Reader's AI Assistant and a number of other AI products, NotebookLM allows you to interact with the sources by typing in a chat box. It also recommends questions you can type to get you started.

I selected a suggested question ("How does AI's influence on education impact ethical considerations?") and received the following reply below. Note that within NotebookLM, each bullet contains one or more numbers that, when clicked, take the user to the section of the source that is being referenced.

AI's growing influence on education raises significant ethical considerations across multiple dimensions. Here's a breakdown of key areas:

1. Bias and Fairness:

• AI algorithms are trained on data, and if that data reflects existing societal biases, the AI can perpetuate and even amplify those biases in educational content and decision-making. This can lead to skewed perspectives, reinforce stereotypes, and result in unfair or discriminatory outcomes for some students.

• For example, text-to-image AI can amplify racial and gender imbalances.

• AI models may not be attuned to shifts in what is considered ethically acceptable, as they rely on historical data.

• It's important that AI outputs are monitored and adjusted to ensure they are appropriate.

• Digital divide bias arises from the underrepresentation of populations with limited digital access, which could further disadvantage those populations.

  
  

2. Privacy and Data Security:

• The use of AI in education involves collecting and analyzing large amounts of student data, raising concerns about confidentiality, privacy, and the potential for misuse of this data.

• AI systems may also be vulnerable to security risks.

• Educational institutions need to ensure compliance with data protection laws and regulations.

  
  

3. Accountability and Responsibility:

• As AI systems become more integrated into education, it is important to consider who is responsible and accountable for their actions, and to ensure they are not used to facilitate harm.

• There needs to be clear mechanisms for addressing errors made by AI systems.

• The question of ‘AI loyalty’ arises: for whom do AI systems work?

• There's a need to clearly define ownership and conflicts of interest related to AI systems.

  
  

4. Academic Integrity and Misconduct:

• The ease with which students can use AI to generate answers raises concerns about plagiarism and academic dishonesty.

• Institutions need to re-evaluate what constitutes academic honesty in the AI era.

• Clear guidelines are needed about what constitutes misuse of AI, for example, using AI to generate essays, while using it to analyze data may be acceptable with proper attribution.

• Over-reliance on AI can diminish independent research and critical thinking skills.

  
  

5. Human Agency and Oversight:

• There is concern that AI could restrict teacher autonomy in pedagogical decisions and that educators may over-rely on automated systems due to automation bias.

• AI should improve the quality of information people use to make informed decisions, rather than making autonomous decisions.

• Maintaining human accountability when making high-stakes decisions will ensure that ethical and responsible actions are taken.

• It is important to value the practices of teachers and researchers when using GenAI.

  
  

6. The Human Touch:

• AI could diminish the human connection in the learning process if face-to-face interaction and feedback sessions are replaced by automated feedback.

• Students may miss the emotional and motivational support offered by human educators.

• There is a concern that important skills like creative thinking and leadership may be overlooked with over-reliance on AI.

• It is important to maintain a balance between leveraging technology and maintaining academic integrity.

  
  

7. Transparency and Explainability:

• It's important that AI systems are transparent in how they work so that educators and students can understand their outputs and limitations.

• Explainable AI is needed so that humans can understand how AI algorithms make decisions.

• There is a need for transparent documentation of design and decision-making processes in AI.

  
  

8. Equity and Access:

• Discrepancies in access to technology and resources could widen the educational divide.

• There is a need to ensure fair and inclusive access to AI resources for all students.

  
  

9. Commercialization:

• The rise of AI in education may lead to conflicts between commercial interests and genuine educational needs.

• There is a risk of companies marketing AI solutions that side-line the learning needs of students.

  
  

10. Impact on the Role of Educators:

• There are questions about the changing roles of teachers as AI becomes more integrated into classrooms, as well as concerns about job displacement.

• AI literacy will be an essential qualification for educators in order to effectively and ethically use AI.

• Educators need to understand the limitations of AI systems and promote responsible use.

• Teachers are essential for promoting ethical AI practices and facilitating discussions about bias, privacy, and inclusivity.

• Teachers should be trained to understand and interpret AI-driven insights.

  
  

11. The Need for Ethical Frameworks and Policies:

• There is a need for clear ethical guidelines and policies for AI use in education.

• These policies need to address issues such as transparency, fairness, accountability, safety, and privacy.

• It is important to monitor and validate AI systems for their ethical and pedagogical appropriateness.

• AI systems should be designed to align with the values of humanity and avoid causing harm.

• AI ethics should not be exploited for economic gain.

• There is a need to avoid "ethics washing," where companies use ethics as a facade to justify deregulation or self-regulation.

• A comprehensive framework that integrates technical and ethical considerations is needed.

• Collaboration between developers, operators, and regulators is needed to craft guidelines and regulations.

• Educational institutions need to create AI-specific guidelines and educate students about ethical and responsible AI use.

• International collaboration is needed for a universal approach to AI regulation.

  
  

12. AI Literacy:

• AI literacy, which includes understanding how AI works, its impacts, and ethical implications, is crucial for both students and teachers.

• AI literacy includes technical skills, critical thinking, and ethical awareness.

• It's important to address the capabilities and limitations of AI in academic study and research.

• Students should be equipped to critically engage with and evaluate information they obtain from AI tools.

  
  

13. Societal Impact:

• AI raises broader societal implications related to ethics, equity, and the economy.

• There are concerns about job displacement due to AI adoption.

• It is important to consider the environmental costs of AI.

  
  

14. Other Ethical Considerations

• AI systems need to be developed to recognize and respond to human emotions.

• It's important to consider how AI can impact human-AI interaction in the workplace.

In summary, integrating AI into education presents both tremendous opportunities and significant ethical challenges. It's crucial to develop and implement AI in a way that prioritizes human well-being, equity, and ethical principles, and to foster an environment where AI is used responsibly and effectively to enhance the learning experience.

Study Guide

I also generated a study guide using the "Study Guide" button. Study guides come with "quizzes" (questions), answers, essays, and a glossary of terms. Although you can selected and deselect your sources, I have found that if you want to focus on a specific source, you are better off loading that source into its own folder. Here is the study guide that NotebookLM generated.

AI in Education Study Guide

Quiz

1. How does the tokenization process in large language models affect the input text?

2. What are the main components of a basic artificial neural network (ANN)?

3. Describe the concept of "jailbreaking" an AI model and provide an example.

4. Explain the difference between generative and discriminative AI models.

5. How do AI systems use "attention" mechanisms when processing language?

6. What is the purpose of the MMLU benchmark?

7. What are some ethical concerns about using AI in education?

8. How can "fine-tuning" improve the performance of large language models?

9. How can AI contribute to sustainable development goals, according to students in the source material?

10. Explain how AI systems can learn from reinforcement learning with human feedback.

Answer Key

1. Tokenization breaks text into smaller units (tokens) like words, parts of words, or punctuation. Changes in punctuation or capitalization can change the tokens created from input text, and the context window restricts how much text can be processed at once.

2. An ANN consists of interconnected artificial neurons. Each neuron multiplies inputs by weights, adds a bias, and applies an activation function to produce an output. These outputs are passed to subsequent neurons in the network.

3. Jailbreaking refers to manipulating an AI model through specific prompts to bypass its safety mechanisms and intended behaviors. An example is the 'DAN' (Do Anything Now) mode for ChatGPT, where users get the model to produce outputs it normally would not.

4. Generative AI models are designed to create new data instances, like text or images, while discriminative models learn to classify data by distinguishing between different categories or predicting labels. Generative models use training data to create new examples, whereas discriminative models are trained to categorize or label existing examples.

5. Attention mechanisms allow AI models to weigh the importance of different words in an input sentence, enabling the model to focus on the most relevant parts to understand the meaning. By assigning weights to words, the model can grasp the essence of a question or prompt.

6. The MMLU (Massive Multitask Language Understanding) benchmark evaluates AI models across 57 tasks from various domains (e.g., math, history, computer science). This tests the knowledge and problem-solving abilities of AI, particularly the abilities of large language models.

7. Ethical concerns in using AI in education include algorithmic bias, inequitable access, over-reliance on AI which may hinder development of higher-order thinking skills, and issues related to student privacy and data security. Also the potential for the homogenisation of thought, over-commercialisation, and misconduct such as plagiarism.

8. Fine-tuning involves further training a pre-trained language model on a smaller, domain-specific dataset to specialize its abilities for specific tasks or industries. This process allows the model to adapt its general knowledge and become more proficient in a particular field, such as legal or medical text.

9. Students envision AI as a solution for environmental problems, including optimizing agriculture, managing water, and cleaning oceans. They believe AI can help change consumption habits, prevent food waste using smart refrigerators and automate greenhouse production.

10. Reinforcement Learning with Human Feedback involves AI models learning from responses rated as helpful or unhelpful by human evaluators. This method can refine AI models' behaviors to align with human preferences, improving the appropriateness and quality of AI generated content.

Essay Questions

1. Discuss the implications of algorithmic bias in AI-driven educational tools, and suggest strategies to mitigate these biases. Consider the sources in your answer.

2. Analyze the potential benefits and drawbacks of integrating generative AI into higher education, focusing on its impact on teaching and learning methods.

3. Explore the ethical challenges associated with using AI in assessment and grading and propose guidelines for responsible implementation, using material from the sources.

4. Evaluate the role of AI in promoting or hindering creative processes in education. Consider the tensions between AI as a support and AI as a replacement of human endeavor.

5. Examine how AI literacy can be incorporated into educational curricula at various levels, from K-12 to higher education, and how this can address the risks and challenges of AI in society, using material from the sources.

Glossary of Key Terms

• Algorithm: A set of instructions or rules that a computer follows to perform a task.

• Artificial General Intelligence (AGI): A hypothetical type of AI that possesses human-level cognitive abilities and can perform any intellectual task a human can.

• Artificial Narrow Intelligence (ANI): Also called "weak AI," this refers to AI systems that are designed and trained for a specific task or problem.

• Artificial Neural Network (ANN): A computational model inspired by the structure and function of biological neural networks, used for machine learning.

• Attention Mechanism: A technique used in AI models that allows them to focus on specific parts of an input when processing information.

• Bias: In AI, bias refers to systematic errors or unfairness in data, algorithms, or outcomes that result in unequal or discriminatory treatment.

• Context Window: The maximum amount of input data a language model can process at once, limiting the size of the input.

• Decoding: The part of the AI process that uses attention and embeddings to predict and create text as output.

• Deep Learning: A type of machine learning that uses multiple layers of artificial neural networks (DNNs) to analyze data.

• Digital Literacy: The ability to understand and use information in multiple formats from a wide variety of sources when it is presented via computers.

• Embedding: A numerical representation of a word that captures its meaning and its relation to other words.

• Encoding: The process of converting data (like words or images) into a numerical format that AI models can process.

• Fine-tuning: The process of further training a pre-trained AI model on a specific dataset for a more specific purpose.

• Foundational Model: A large AI model trained on a huge amount of data which can be fine-tuned for more specific uses.

• Generative AI: A type of AI that can generate new content such as text, images, audio, or video.

• Jailbreaking: The act of manipulating an AI model to bypass safety mechanisms to produce unintended outputs.

• Large Language Models (LLMs): A type of AI model with billions of parameters and trained on vast amounts of text.

• Machine Learning: A type of AI that allows computers to learn from data without being explicitly programmed.

• Reinforcement Learning with Human Feedback (RLHF): A method for training AI models that uses human evaluation to help them improve.

• Self-Attention: A mechanism that allows a model to weigh the importance of different words in an input sentence

• Tokenization: The process of breaking down text into smaller pieces (tokens) that an AI model can process.

• Vector: A numerical representation of data (like text or images).

• Weight: A numerical value that indicates the strength of a connection between neurons in a neural network.

Sharing

Notebooks in NotebookLM can be shared with others, either with view only or editing permissions. The collaborator must provide a Gmail e-mail in order to access the notebook, however. Users of the paid versions have additional sharing options.

What's Next?

I have only touched the surface of what you can do with NotebookLM, but my purpose with this post is not to hype this product, but rather to show the capability of the AI products that are already available. NotebookLM was selected because it is currently an obsession of mine, largely due to the audio overview feature (disclaimer: the author has no association with Google Labs and no financial interests in NotebookLM). It is also free for anyone with a Google Account and very easy to use. If you haven't engaged much with AI applications, NotebookLM may be a good place to start. Upload a source you are familiar with, create an audio overview, use the chat, and explore what NotebookLM does well, and what could be improved. In future posts we will examine some other tools that are available as well.

In the third post of this introductory series, now that we have defined our terms and described some of the current uses of AI in education, we will discuss some of the opportunities and challenges AI presents.

News and Announcements

Interested in learning more about AI? 

Regis faculty are encouraged to self-enroll in the "Using AI in your Teaching" course created by Dr. Diana Perdue. Faculty can access the course here (don't forget to click the "enroll me" button).

Interested in contributing a post or joining our Regis College Artificial Intelligence Faculty Learning Community?

Contact Allen Karsina at allen.karsina@regiscollege.ed

---

References

The sources loaded into the notebook in NotebookLM at the time this post was written are included below. I asked NotebookLM to generate the reference list for me. It gave me a nice starting point but there were many errors - another reminder that the output of AI must be critically analyzed at all times.

Baker, R. S. & Hawn, (2020). Algorithmic Bias in Education. Retrieved from https://learninganalytics.upenn.edu/ryanbaker/AlgorithmicBiasInEducation_rsb3.7.pdf on December 25, 2024.

Bobula, M. (2024). Generative Artificial Intelligence (AI) in higher education: a comprehensive review of challenges, opportunities, and implications. Journal of Learning Development in Higher Education, 30, 1-27.

Chan, C. K. Y., & Colloton, T. (2024). Generative AI in higher education: The ChatGPT effect. Routledge.

Giarmoleo, F. V., Ferrero, I., Rocchi, M., & Pellegrini, M. M. (2024). What ethics can say on artificial intelligence: Insights from a systematic literature review. Business and Society Review, 129(2), 258–292. https://doi.org/10.1111/basr.12336

Jennings, A. M., & Cox, D. J. (2024). Starting the conversation around the ethical use of artificial intelligence in applied behavior analysis. Behavior Analysis in Practice, 17(1), 107–122. https://doi.org/10.1007/s40617-023-00868-z

Kim, S. L., Cheong, M., Srivastava, A., Yoo, Y., & Yun, S. (2020). Knowledge sharing and creative behavior: The interaction effects of knowledge sharing and regulatory focus on creative behavior. Human Performance, 34(1), 49–66. https://doi.org/10.1080/08959285.2020.1852240

Lee, S. S., & Moore, R. L. (2024). Harnessing Generative AI (GenAI) for Automated Feedback in Higher Education: A Systematic Review. Online Learning Journal, 28(3), 85-104. https://doi.org/10.24059/olj.v28i3.4593

Parasuraman, R., & Manzey, D. H. (2010). Complacency and bias in human use of automation: an attentional integration. Human Factors, 52(3), 381-410. DOI: 10.1177/0018720810376055

Resnik, D.B., Hosseini, M. The ethics of using artificial intelligence in scientific research: new guidance needed for a new tool. AI Ethics (2024). https://doi.org/10.1007/s43681-024-00493-8

Urmeneta, A., & Romero, M. (2024). Creative Applications of Artificial Intelligence in Education. Palgrave Macmillan.

West, D., Luzeckyj, A., Toohey, D., Vanderlelie, J., & Searle, B. (2020). Do academics and university administrators really know better? The ethics of positioning student perspectives in learning analytics. Australasian Journal of Educational Technology, 36(2), 60–70. https://doi.org/10.14742/ajet.4653

Yang, S. J. H. (2021). Guest Editorial: precision education - a new challenge for AI in education. Educational Technology & Society, 24(1), 105–108.

Acknowledgements 

ChatGPT 4o was used to suggest improvements for the text for some portions of this post.

This post was written by the author with the exception of the excerpts demonstrating the excerpts from NotebookLM.

The image was generated using a GPT (AI Art Assistant) created by the first author. The author used the prompt: futuristic laptop, black background, complex cosmic equations, existential expressionism, ink splatter, black and white.

The audio was generated using NotebookLM as described in the post.