Greetings and salutations — welcome to another newsletter!
Whatever our personal feelings are about AI, the technology is here to stay — and advancing at a good clip. This newsletter will discuss a few challenges and opportunities for learning with AI, then give a few general recommendations for learning with AI.
Ideas
Here is Belcher’s Model for Learning — a short refresher:
- Our body and brain receives information from our senses, with the brain selecting information on which to focus.
- One way to think about processing in the brain is with two parts: Working memory and long-term memory. Working memory combines information from sensing and focusing with prior knowledge and skills in long-term memory, then uses the combination to do something with the knowledge and skills. Long-term memory is the “storehouse” of knowledge and skills, which you are continually modifying.
- The output of the learning process is practicing and performing, which is using the knowledge and skills in some way.
- (For a much deeper discussion about the entire Model for Learning, read this essay: That’s How Learning Works?!?! A Comprehensive Model for Understanding the Learning Process.)
If we are serious about learning a set of knowledge and skills, then we need to do a few actions with the knowledge and skills:
- Look at the knowledge and perform the skills enough to have fluent recall — the ability to answer questions about the knowledge and skills quickly and easily (strong storage in long-term memory).
- Have an understanding of the knowledge and skills — instead of simply memorizing the names for different parts, understanding is the ability to know the knowledge and skills at deeper levels (strong storage in long-term memory).
- Create connections for many different parts of the knowledge and skills, which organizes the knowledge and skills (strong storage in long-term memory) — this increases your ability to use the knowledge and skills in standard and new situations (more information in working memory).
[These three bullets can be combined into an acronym: FRUCO (fluent recall, with understanding, connections, and organization).Thanks to one of my professors, Dr. Paul Heideman, through his ideas on learning!]
Challenges for Learning with AI
As with any type of tool, there are challenges for learning with AI.
The biggest challenge is relying too much on AI for understanding, connections, and organization.
Learning is hard — there is a ton of time and effort required to create understanding, connections, and organization for knowledge and skills. By having the AI “do” this for us, our brains do not have the chance to physically and conceptually learn knowledge and skills. Taking the time and effort to learn is especially critical for those who are just beginning; although beginners may want to skip the time and effort, skipping usually causes big gaps in their knowledge and skills.
To be fair to AI, this is not a new phenomena — I saw this all the time as a high school physics teacher. Although most students have done basic physics before, learning physics requires time and effort. I created the conditions for students to understand, make connections, and organize information through a bunch of different ways, but ultimately the student had to put in the work. This is the same for serious study in every field; we need to discuss this with learners at the beginning stage.
Another challenge is the potential for AI to provide incorrect or biased information.
The information for sets of knowledge and skills are stored in long-term memory, so that information needs to be correct. If the information is incorrect, then our understanding is also incorrect — leading to major problems when we try to apply knowledge and skills. Although the AI models are advancing, there are still issues with incorrect information; these issues can cause you to learn the wrong information.
Another part of this is bias: On what information was the AI trained? Many fields have ideas that seem “correct” right now, but these ideas are actually just opinions that are useful in a small number of situations. There are also cultural and societal issues in the information used to train the AI, which can introduce more bias into the information you are learning.
Opportunities for Learning with AI
Given the major challenges, there are opportunities for learning with AI.
AI can give rapid feedback, which can increase the rate of learning.
Though there are nuances to the amount and timing of feedback, learning definitely requires feedback. As you practice or perform a set of knowledge and skills, there will be an outcome — your outcome can be compared to the expected or desired outcome. Before AI the feedback process was (typically) done by a teacher or coach, but the feedback process can now be done by AI.
The positive part about AI is the on-demand nature; human teachers or coaches have to sleep or be somewhere else, so they were not always available. With AI you can get feedback whenever the feedback is needed, which can satisfy our curiosity and continue motivation. However, the comparison of outcomes and error analysis for the outcomes needs to be checked — the AI may or may not actually know why the error was made, so we have to be careful about drawing conclusions.
AI can create novel and detailed scenarios that simulate a performance environment.
Related to the feedback is the ability of AI to create novel and detailed scenarios, which can be used in many fields and in many ways. One interesting use of this was in an article on Jayden Daniels, who is a rookie quarterback for the Washington Commanders. He completed one of the best seasons for a rookie quarterback in the history of the NFL; part of his preparations before a game was to put on a virtual-reality headset that simulated what he would see in the actual game later in the day. My theory is that this simulation before the game helped Jayden learn quickly, giving him confidence to perform during the actual game.
Simulations have a long history of uses, ranging from simple paper-and-pencil games to complete immersive environments. Performing a simulation with AI and then chatting with AI about parts of the simulation — using reflective questions — can help with every part of the learning process: Embedding information in long-term memory; creating connections and better organization in long-term memory; and, using the knowledge and skills with sensing and focusing in working memory.
AI can make far-reaching connections.
Experts are experts because they have a ton of detailed, strongly connected, and highly organized sets of knowledge and skills for a specific field. However, experts still have their blind spots; AI can help with these blind spots by making far-reaching connections. We have seen this prominently in the chess world: The AI chess players can run massive amounts of computations, which allows them to make extremely far-reaching connections. Expert human chess players work with AI chess players to understand these connections and use the connections in their own playing. With the additional capabilities of general AI, experts in many fields are using the general AI to make far-reaching — and unexpected! — connections in their field and between fields. This is exciting for these experts because the connections will help drive research forward by giving new pathways for exploration.
General Recommendations
Here are a few general recommendations for learning with AI:
- Do the hard work of thinking as you use AI — there are no shortcuts to the time and effort required to become an expert in a field.
- Double-check the information for accuracy and bias.
- Use AI as a feedback tool — comparison of outcomes and error analysis.
- Create and use simulations to perform in novel situations, then analyze the process during and outcomes from the simulation.
- If you are an expert in a field, use AI to make far-reaching connections.
Questions
- How does my Model for Learning combine with ways to learning using AI?
- What do you believe to be the point of learning?
- Do you disagree or agree with fluent recall, with understanding, connections, and organization?
- Do you disagree or agree with the challenges?
- What are other challenges for learning with AI?
- Do you disagree or agree with the opportunities?
- How do you use AI to learn?
Learning happens when we share what we are thinking, so I would love to hear your answers! Also, you can use these questions as conversations starters with friends and family — hearing their answers would be great!
If this newsletter resonated with you, please share on the socials and with someone who you think would also benefit; I would greatly appreciate any help in spreading these ideas!
Thanks for reading this newsletter — and all the best!
Nathan
