AI, help me with a difficult reading
A learning practice from one of my students
(NB: for some posts I’m going to place parts of them behind paywalls. I love doing this research and hope it is useful to readers, but it takes a growing amount of my time and, as an independent researcher, I need to make this project sustainable.)
How might we use generative AI to enhance learning?
This is a question I’ve been exploring intensely for a while. I’ve brought it into my Georgetown University classes during this semester, to see what students discover and think. Last week one of them invented a use of AI which struck me as clever, and which he gave me permission to write about.
Sicong Ma was curious about how generative AI might help him make progress through difficult readings. He came up with a fascinating prompt, asking ChatGPT to “generate hypothetical scenarios” for a text, then copied and pasted in a sample. I think he started with Walter Benjamin’s classic “Work of Art in the Age of Mechanical Reproduction” (1935) (here’s one copy), which was the second-hardest reading in our Technology and Innovation seminar.
Here’s the prompt I used, based on Sicong’s work: “Please help me understand the following passage. Provide hypothetical scenarios to make better sense of it,” followed by most of sections III and IV.
Here’s how ChatGPT responded. It started with a quick introductory sentence, which set the stage: “This passage discusses how human perception and the value of art have evolved over time, influenced by historical, social, and technological changes.”
Then on to the main answer:
Here are a few hypothetical scenarios to help clarify these concepts:
Changing Modes of Perception:
Ancient Times: Imagine a society where the only way to experience art is by physically visiting it. The perception of art is influenced by the effort to reach it and the rarity of the experience.
Modern Times: Now, consider our world where digital images of art are instantly accessible online. The way we perceive and value art has changed because it's more easily accessible and widely reproduced.
The Concept of Aura:
Original Art: Think of viewing the Mona Lisa in the Louvre. The aura is the unique experience of seeing the original painting in its historical and cultural context.
Reproduction: Now imagine looking at a postcard of the Mona Lisa. The aura is diminished because the postcard is a mass-produced item, lacking the unique context and experience of the original.
Art and Ritual vs. Art and Politics:
Ritualistic Art: In ancient civilizations, art often served a ritual or religious purpose. Consider a tribal mask used in ceremonies. Its value is tied to its role in these rituals.
Art and Politics: In contrast, consider a political poster in today's society. Its value lies in its political message and its ability to be mass-produced and distributed, not in any ritualistic or religious function.
Art for Art's Sake:
Traditional Art: Imagine a Renaissance painting, valued for its beauty and craftsmanship.
'Pure' Art: Now think of a modern abstract painting, valued not for representing anything in the real world, but for its intrinsic qualities as art, independent of social or political context.
Mechanical Reproduction's Impact:
Before Photography: A sculptor creates a statue. There's only one original piece, and its value lies in its uniqueness and the skill required to create it.
After Photography: A photographer takes hundreds of photos, which can be endlessly reproduced. The value shifts from the rarity and uniqueness of a single piece to the content of the image and its ability to be widely disseminated.
Finally, the bot concluded with a summary, tying the examples back to its expression of the passage’s overall theme: “These scenarios illustrate how changes in technology, society, and politics have altered the way we perceive, value, and interact with art, moving from a focus on unique, ritualistic items to mass-produced, politically charged works.”
I’m struck by several things in this result, starting by how ChatGPT didn’t create scenarios for the text as a whole. Instead, it broke the response into several points, then generated hypothetical scenarios under each, further breaking those out into contrasting sub-categories: before and after, ritual vs politics, and so on. In terms of style, I note how it uses imperative language for almost every example: imagine, think, consider. Overall, it did a good job of explicating the passage.
Still, I wanted more. As a teacher I often work students’ lives and interests into discussion, especially for creating examples. So I asked Chat GPT this: “Thank you. Can you revise your answer to offer more personalized scenarios?”
Here’s the response:
Certainly! Let's personalize these scenarios to better relate to your interests and context:
Changing Modes of Perception in Higher Education:
Traditional Classroom: Imagine teaching or attending a class in a traditional university setting. The experience and perception of learning are shaped by the physical presence, the classroom environment, and the interaction with peers and instructors.
Online Learning: Now, think about the shift to online education, perhaps accelerated by recent global events. The way students and educators perceive and engage with educational content has transformed. Digital platforms, virtual classrooms, and asynchronous learning have become the new norm, altering the traditional educational 'aura.'
The Concept of Aura in Your Home:
Unique Artifacts: Consider a unique, handcrafted piece of art or a family heirloom in your home. Its value and the perception of it are tied to its uniqueness and personal history.
Mass-Produced Items: Contrast this with a commonly available, mass-produced decorative item. Though it may be aesthetically pleasing, its 'aura' is lessened due to its ubiquitous nature and lack of personal history.
Cultural Shifts in Manassas, Virginia:
Historical Perspective: Manassas, with its rich Civil War history, offers a unique perspective. Imagine engaging with historical sites or artifacts in your area. Their significance is rooted in their historical context and unique story.
Modern Interpretation: Now think about how these historical elements are presented in modern media, like documentaries or online articles. The mode of engagement changes, offering a different, perhaps less immersive, experience than physically being at a historical site.
Art and Personal Ritual:
Personal Rituals: Reflect on a personal ritual that involves art, like visiting a favorite local museum or gallery. The experience is unique and personal, creating an 'aura' around the artworks you encounter.
Digital Art Viewing: Compare this to viewing art online, perhaps through virtual museum tours. While still enriching, the experience lacks the personal ritualistic element, altering the perception and value of the art.
Impact of Technology on Personal Life:
Pre-Digital Era: Consider a time when the only way to view family photos was through physical albums. Each photo was unique and held a special value.
Digital Age: Now, with digital photography, you can take, store, and share countless photos. While this is convenient, the individual photos may lose some of their unique emotional value due to the sheer volume and ease of reproduction.
These personalized scenarios aim to connect the concepts from the passage to aspects of your life in higher education, your local history and environment in Manassas, and the impact of technological changes on personal experiences.
Wow. This brought my life right into the discussion, including my professional research into higher education and the location of my home in northeastern Virginia. ChatGPT also shifted language to address me in a more generic but still personal way, asking me to think of my home and my personal habits.
“Work of Art” is a classic humanities essay. How might Sicong’s method work on the sciences?
As a test case, I picked a controversial new paper in climate science, “Global warming in the pipeline,” by James Hansen et al. It’s published in an open access journal, so you all should have access to it. (For a sample of the controversy, here’s Michael Mann’s critique) I picked what I thought of as one of the crucial sections. It begins under the header “LGM-Holocene and PGM-Eemian evaluation of ECS” and sets out to do the following:
In this section we evaluate ECS by comparing neighboring glacial and interglacial periods when Earth was in energy balance within less than 0.1 W/m2 averaged over a millennium. Larger imbalance would cause temperature or sea level change that did not occur [48].4 Thus, we can assess ECS from knowledge of atmospheric and surface forcings that maintained these climates.
Here’s my prompt: “Please help me understand the following passage. Provide hypothetical scenarios to make better sense of it, including personal examples,” followed by a long copy-and-paste.
ChatGPT begins its response as before, with a one-sentence summary followed by a one-sentence introduction of what it was going to do:
This passage is about recent advances in estimating past climate conditions, specifically focusing on the Last Glacial Maximum (LGM) and its associated paleotemperatures, greenhouse gas (GHG) forcings, and the concept of Equilibrium Climate Sensitivity (ECS). Let's break it down with some hypothetical scenarios and personal examples:
Advanced Analysis Techniques for Paleo Temperatures:
Scenario: Imagine you're a scientist using advanced tools to analyze historical weather data in Manassas, Virginia. Earlier methods might have relied on limited physical records, but now you can use sophisticated computer models and a broader range of data, like tree rings or ice cores, to get a more accurate understanding of past climate patterns.
Geochemical Proxies and GCMs:
Personal Example: Think about using a thermometer to measure temperature vs. using a complex weather app that takes into account various data points like humidity, wind speed, and historical patterns to predict the weather. Similarly, scientists use geochemical proxies (like isotope data) and General Circulation Models (GCMs) to estimate past sea surface temperatures, offering a more nuanced understanding than previous methods.
Cooling During the LGM:
Hypothetical Scenario: Picture Manassas during a particularly cold winter, where temperatures drop significantly compared to the average. This is akin to the cooling during the LGM, except on a much larger and longer scale, with global temperatures dropping by about 7°C.
Forcing and Climate Sensitivity:
Personal Example: Imagine adjusting the heating in your home. The amount of heat you add or remove (forcing) affects the indoor temperature (climate response). Similarly, changes in GHG levels and surface characteristics like ice sheets and vegetation during the LGM had a "forcing" effect on Earth's climate. Scientists calculate ECS to understand how much global temperature would change in response to these forcings.
Human Influence in the Latter Holocene:
Scenario: Consider how your activities in Manassas, like driving a car or using electricity, contribute to changes in atmospheric composition. This is a small-scale example of how human activities in the latter Holocene have started influencing global GHG levels and aerosols, impacting climate.
GCMs and Clouds in Climate Modeling:
Hypothetical Scenario: Imagine trying to predict how much sunlight your many cats will enjoy in your living room. You need to consider not only the position of the sun but also cloud cover. Similarly, GCMs need to accurately represent clouds to evaluate climate sensitivity and the effects of surface albedo changes.
It concludes as before, with a summary of what it tried to offer:
These scenarios help to contextualize the complex processes discussed in the passage, relating them to familiar situations and your personal environment. The passage essentially highlights the sophistication and challenges in understanding and modeling past climate changes, emphasizing the role of various factors in influencing Earth's climate.
The personal dimension is pretty solid, addressing my geographical location, asking me to think of many aspects of my life, and inviting me to do stuff (imagine, consider, picture). It keeps tying those personal details to larger ideas, including climate change as a whole.
I felt these hypotheticals might have been too easy, so I asked for a more challenging version: “Thank you, but can you make the hypothetical scenarios more advanced and challenging, please?”
Its response: “Certainly! Let's delve into more advanced and challenging hypothetical scenarios to understand the passage:”
Advanced Techniques in Paleo-Climate Reconstruction:
Advanced Scenario: Imagine a high-tech laboratory where scientists use cutting-edge molecular techniques to analyze ancient ice cores. They're not just looking at the physical characteristics of the ice but analyzing isotopic compositions to infer minute details about past temperatures. This is akin to using forensic analysis to solve a historical mystery, where every molecular clue sheds light on the Earth's climatic past.
Geochemical Proxies and GCMs in Detail:
Complex Example: Consider a scenario where you're a climate scientist working with a supercomputer. You're inputting data from various geochemical proxies, like isotopic ratios from marine sediments, into a General Circulation Model. This model simulates Earth's climate system with high complexity, accounting for atmospheric and oceanic currents, ice sheet dynamics, and even vegetation changes. Your task is to reconstruct how the sea surface temperature (SST) changed during the LGM, a puzzle requiring integration of complex data sets and advanced modeling.
Understanding LGM Cooling with Modern Analogues:
Challenging Scenario: Visualize leading a research expedition to Antarctica. Your goal is to understand the cooling during the LGM by studying present-day glacial dynamics and temperature patterns. You collect ice samples, measure current temperature variations, and compare these with historical data and models. This hands-on approach in an extreme environment mirrors the challenges scientists face in reconstructing past global cooling events like the LGM.
Forcing and ECS in Climate Studies:
In-depth Analysis Scenario: Assume you're analyzing the impact of deforestation in the Amazon on global climate. You measure changes in local albedo (reflectivity) due to deforestation and model how these changes contribute to global temperature changes (forcing). By comparing these data with historical GHG levels and temperature records, you're essentially conducting a scaled-down version of what scientists do to calculate Earth's Equilibrium Climate Sensitivity (ECS) - understanding how sensitive Earth's climate is to changes in greenhouse gas concentrations and surface characteristics.
Holocene Human Influence and Aerosol Forcing:
Sophisticated Scenario: Picture yourself as an environmental policy analyst assessing the impact of industrialization on climate change since the Holocene. You're examining data from various sources, including satellite imagery of aerosol concentrations, industrial emission records, and deforestation rates. Your analysis aims to discern the human-made contributions to climate change, distinguishing them from natural variations, much like scientists separate natural aerosol changes from human-made effects in climate studies.
Challenges of Cloud Representation in GCMs:
Complex Modeling Scenario: Imagine you're a software engineer developing a new module for a GCM that accurately simulates cloud formation and behavior. This task is crucial because clouds significantly affect Earth's energy balance. Your module must account for various cloud types, their interactions with solar radiation, and their impact on surface temperatures. This complex modeling reflects the real challenges scientists face in incorporating cloud dynamics into climate models, as mentioned in the passage.
These scenarios represent more intricate and challenging situations that parallel the complexities and methodologies discussed in the passage, offering a deeper understanding of the advanced scientific processes involved in climate research and modeling.
Better. I fear the results talk more about being challenging than they actually are, but overall it is more advanced than the first try.
Overall, I think Sicong Ma has come up with a very useful way to get generative AI to help learners with challenging readings. Please feel free to try this yourselves. What do you make of it?