Education with Generative AI

The rise of generative AI technologies like ChatGPT-4 currently paves the way toward innovation of pedagogical approaches. Examples of revolutionizing the learning process include courses on academic writing where students are given writing prompts by an AI module which then produces a piece of text that resembles human writing, or courses on programming where students are given autocomplete-style suggestions as they code, or by writing a natural language comment describing what the code should do.
Original content like text, images, or videos can be generated by generative AI tools, which have been trained on massive online datasets and are increasingly being integrated into various office tools and educational platforms. Major software companies, such as Microsoft, plan to incorporate generative AI in all its tools, including Word, Excel, and its Bing search engine.

Generative AI tools for education

The report “innovating pedagogy 2023” and personal experiences suggest a list of tools which are currently available, not free, but useful for teaching and learning, include:

• https://www.decktopus.com/ helps produce professional presentations.
• https://monica.im/ summarizes articles (in Chrome).
• https://stockimg.ai/ generates images for your project.
• https://openai.com/product/dall-e-2 creates realistic images from a text
  description.
• https://www.chatpdf.com/ upload a pdf and ask it questions.
• https://sheetplus.ai/ translates text into spreadsheet formulae.
• https://agentgpt.reworkd.ai/ breaks task into steps and works through them.
• https://books.google.com/talktobooks/ ask questions of Google Books.
• https://bearly.ai/ summarizes long papers
• https://www.perplexity.ai/ short answers to questions, with references.
• https://www.questgen.ai/ generates quizzes from text.
• https://tldrthis.com/ summarizes any article.
• https://www.heygen.com/ create videos from text in minutes with AI-generated avatars and voices.
• https://www.aomni.com/ finds, extracts and processes Internet data.

The rapid broadening of the spectrum of applications for generative AI in education is noticeable. The options of a student using an AI tool to automatically summarize dense research articles, or a group project team leveraging AI as a collaboration coach are already reality. Learning assistance, team activities facilitation, understanding improvement, and data exploration accessibility can be offered by these tools, from personal life coaches to study companions or co-workers. If you are looking for other tools, check https://theresanaiforthat.com.

Cultivating AI Literacy: An Urgent Need

The application of AI in education goes beyond content creation. For example, conversational AI chatbots can benefit language learners by allowing them to practice pronunciation, similar to how voice assistants such as Siri and Alexa process spoken language. Similarly, an AI tool that adapts learning materials based on individual needs could provide a student with personalized study guidelines, reminiscent of how AI can provide tailored educational content for improved learning. Furthermore, in the area of metacognition, AI-driven reflective tools can prompt students to analyze and understand their thinking patterns, such as cognitive bias detectors, thus fostering deeper self-awareness. Supporting language acquisition, personalized tutoring, assisting with cognitive processes such as decision-making through data analysis, and memory enhancement through AI-powered reminders are further examples of the transformative potential of AI in education. In this process, it is useful to understand how to best generate the input to an AI tool to generate an output that is of value.

Mastering writing through prompt engineering

Prompting, or prompt engineering, is a strategy of carefully designing the input, or “prompt”, to a language model such as GPT-4 in order to get the desired output. By refining the questions and statements posed to the model, researchers and students can obtain more scholarly, accurate, and contextually appropriate responses.

Examples:

Literature Review Assistance:

• General Prompt: “Tell me about studies on climate change.”
• Engineered Prompt: “Provide a summary of peer-reviewed studies published in the last five years that discuss the impact of climate change on polar bear populations.”

Thesis Statement Formulation:

• General Prompt: “I need a thesis about renewable energy.”
• Engineered Prompt: “Construct a thesis statement that examines the economic advantages of transitioning to renewable energy sources in developing countries between 2000 and 2020.”

Data Interpretation Guidance:

• General Prompt: “How to analyze data?”
• Engineered Prompt: “Provide a step-by-step methodology for conducting a regression analysis on data examining the relationship between air pollution levels and respiratory illnesses in urban areas.”

The goal is to provide enough information and context in the prompt, iterate and adjust it to get the model to produce the intended response. By tailoring the prompts given to AI in these ways, academic writers can ensure that the information or assistance they receive aligns better with scholarly standards of their research domain.

Concerns on bias, ethics and scientific integrity

However, pitfalls associated with these tools cannot be ignored. Potential challenges related to social biases could be unveiled in situations where a piece of culturally insensitive text is produced by the AI, due to inherent social biases in its training data. Using large language models for purposes, such as fact checks, is another challenge when references are not included. Answers from such tools can seem so convincing that students forget to check basic facts, and just trust falsely the generated content. Do the works cited to in the tool actually exist? Are the biographical facts correct, or are some of these simply “hallucinations”? Does the tool always assume that the experts are white men?
Novice users and students need guidelines to help them understand the kinds of problems that can arise and the answers they get.

A particularly concerning ethical dilemma arises when students rely heavily on these tools to generate complete solutions for assignments without truly grasping the underlying concepts or methodologies. This not only undermines the very purpose of education but also hampers the holistic development of a student’s cognitive and problem-solving abilities.

Such over-reliance on technology can lead to a superficial understanding of subjects, leaving students ill-prepared for real-world challenges where they might not have immediate access to AI aids. Furthermore, it raises questions about academic integrity and the true value of earned grades and qualifications.

Educators, while appreciating the benefits of AI, must be vigilant. Strategies to alleviate this concern might include:

1. Integrated Assessments: Designing assignments and tests that focus on a student’s analytical and critical thinking abilities rather than rote or procedural knowledge.
2. Reflective contributions to AI: Conducting reflective contributions that emphasize the ethical use of AI tools in academics, making students aware of the long-term consequences of over-reliance.
3. Tool Limitation: Implementing AI tools that assist rather than complete tasks. For instance, a tool might provide hints or frameworks rather than full solutions.
4. Honor Codes: Reinforcing academic honor codes that specifically address the use of AI tools, ensuring that students understand the boundaries.

In essence, while AI can be a potent ally in the educational journey, striking the right balance between assistance and independence is crucial to preserving the sanctity and objectives of the learning process.

A shift in teaching and assessment practices

The necessity of training educators and students to use these tools effectively becomes crucial as these technologies permeate the education system. Compared to the introduction of electronic calculators 50 years ago, when mathematics teachers had to rethink what and how to assess, generative AI requires rethinking assessment in every subject area. The impact of AI-generated content on conventional assessments must be imperatively included in the design of future courses. There is a need to move to different forms of assessment, including complex tasks, real-world tasks, reflections on the process of answering a task and critiques of the answers produced by the AI. This is more necessary than ever to avoid degrading education to a mock-up where students simulate competences in line with learning objectives instead of developing them.

In summary, with the emergence of popular AI tools, the transformation of education is necessary and there are not only major challenges and concerns, but also opportunities to be considered.