On rainy days physical science laboratory ten, which measures the speed of sound, is done using resonance tubes and tuning forks. On days without rain the laboratory is done outside by timing the delay between seeing two-by-fours clap and hearing the clap across distances ranging from 300 to 500 meters. The class does one or the other, but never both. This term the class performed the indoor resonance tube version of the laboratory.
A laboratory report was submitted that mentioned having done both versions of the lab and included vocabulary never mentioned in class. This suggested that the student may have copied and pasted the entire directions into a large language model system, neglecting to delete the outdoor version of the laboratory from the input.
The following looks at what happens when raw data, a graph, and directions without editing are uploaded into a large language model, ChatGPT.
Screenshots of the data table and graph were uploaded into ChatGPT 4.0 mini.
ChatGPT ran optical character recognition on the screenshots, capturing even the subscripts.
The laboratory directions were then uploaded as a copy and paste from the Moodle assignment, which includes brief descriptions of both the indoor and outdoor version of the laboratory. No editing of the directions was done.
The two variants are seen above.
The report structure was included in the uploaded text. ChatGPT understood the task fully.
ChatGPT then assembled a solidly done laboratory report missing only one element - the speed of sound based on the inverse relationship displayed on the graph.
The report included references to both versions of the laboratory. For reference the ChatGPT laboratory report is included below. The report is actually well done, minus the one missing element. That was addressed in a follow-up question to ChatGPT and ChatGPT then produced the missing result, although not by using a best fit inversion regression.
The report as written by ChatGPT:
Laboratory Report: Investigating the Speed of Sound
Introduction
In this laboratory experiment, we investigate the nature of sound and its propagation. Specifically, we aim to determine the mathematical relationship between time and distance for sound waves, define sound itself, and calculate the speed of sound based on experimental measurements. We will also explore whether the speed of sound varies with increasing distance when measured outdoors and compare our experimental results to the known published value. For the indoor variant, we will determine the wavelength and frequency for multiple tuning forks, using the resonant distance to calculate the speed of sound.
Data Table
Below is the collected dataset showing the measured wavelengths (ℓ_1 in cm) and their corresponding frequencies (f_1 in Hz). A full dataset screenshot from Desmos is included, with labeled variables and units.
(Insert Desmos screenshot of the table here)
Data Graph
A graphical representation of the collected data is provided below. The axes are labeled as follows:
- X-axis: Wavelength (cm)
- Y-axis: Frequency (Hz)
- A fitted trendline represents the inverse relationship between frequency and wavelength, as expected from the wave equation.
(Insert Desmos screenshot of the graph here)
Data Analysis
From the dataset and graph, we observe an inverse relationship between frequency and wavelength, confirming that the speed of sound follows the equation:
Using the collected data, we compute the speed of sound by taking the product of the measured wavelengths and their corresponding frequencies. The experimentally determined speed of sound is compared to the standard published value (~343 m/s at room temperature).
(Insert Desmos screenshot of calculated values here)
For the outdoor variant, we analyze whether the speed of sound changes with increasing distance. The dataset suggests that the speed remains relatively constant, supporting the principle that sound propagates at a consistent velocity under similar atmospheric conditions.
For the indoor variant, we calculate the wavelength of each tuning fork as four times the resonant distance and use the measured frequency to determine the speed of sound. This allows us to verify our results against known values.
Discussion and Conclusion
This experiment successfully explored the properties of sound and its propagation. The inverse relationship between wavelength and frequency was confirmed through graphical analysis. The calculated speed of sound closely matched the published value, with minor discrepancies likely due to environmental conditions and measurement precision. Additionally, outdoor measurements indicated that the speed of sound does not significantly vary with increasing distance under stable conditions. Indoor measurements using tuning forks provided another verification method for determining the speed of sound.
Overall, this laboratory investigation provided a comprehensive understanding of how sound travels and how its speed can be experimentally determined.
~~~
As follow-up, ChatGPT was asked to provide the speed of sound.
Because ChatGPT chose to calculate the average based only on the first eight of twenty-two data values, and the two most extreme values were the last two of the 22 data values, ChatGPT obtains a significantly different result from the inverse regression.
The inverse regression takes into account all 22 data points with the last two being (24 cm, 1024 Hz) and (208 cm, 149.5 Hz). This results in an estimated sound speed of 28,441 centimeters per second.
The primary goal in this general education physical science for non-majors is not necessarily obtaining the most accurate result possible, the primary goal is to establish in the minds of the students that physical science is based on measurable data and the mathematical models that arise from that data. Results are not a matter of opinion, they are measurements that can be replicated. Refined, improved, perhaps, but facts. This course considers science as an
emergent truth. For the students, the speed of sound is what their data tells them is the speed of sound. The history of science is that of initial measurements being constantly refined as better methods of measurement and improved measurement technologies.
The laboratory reports then document the findings of the student scientists. With a primary goal of "doing science," the documentation is essentially an after action report. What tools should be permitted to produce this report? What tools will these students have available to them over the next forty years?
Students have to be prepared for tomorrow. Reports are no longer written by hand with hand drawn diagrams in pen and ink. Nor are they done on a manual typewriter with graphs printed by a plotter, cut out with scissors, and pasted onto a page. IBM Selectric typewriters included lift-off tape that removed typed letters upon backspacing. Word processing software on computers first introduced spell checking and then grammar recommendations. At present word processing software flags spelling errors and makes grammar recommendations.
Professional writers benefit from having an editor review their material. Google Gemini now provides those services.
A piece from a couple of years ago was copied into Google docs:
The tree was still there. I had scrambled up behind Walung elementary school following tracks that led only to a bamboo grove. From the bamboo I made my way up to the ridgeback. If there is a passageway in the forests of Kosrae, then they are to be found atop the ridge. The shadowy figure of a dark concrete water catchment tank could be made out.
Just past the tank, a cluster of dieffenbachia sprouted from the forest floor. I was not alone – the dieffenbachia assured me someone had been here before. I had been here before. As I was now.
Beyond the dieffenbachia I stumbled across the coral paver walkway. Literally. The walkway was obscured under leaf litter and ground growth, the edge still able to catch my foot. Filtered sunlight dappled the sea of thelypteris ferns around me. A Micronesian dove cooed in the distance.
I followed the trace of the coral pavers up the ridgeline, as I had some 30 years earlier. Where the pavers ended, however, the tree was not on my left as my memory had expected. I looked around, but could not see the tree. Turning right, I headed further up the slope.
Once the piece was copied into Google Docs, Refine with Gemini appeared at the bottom of the text.
Clicking on Refine brought up a list of options. Edit was not an explicit option, but Gemini is a large language model so requests can be made of the system.
The edit command produced the following suggested text.
| Original | Edited by Gemini |
The tree was still there. I had scrambled up behind Walung elementary school following tracks that led only to a bamboo grove. From the bamboo I made my way up to the ridgeback. If there is a passageway in the forests of Kosrae, then they are to be found atop the ridge. The shadowy figure of a dark concrete water catchment tank could be made out.
Just past the tank, a cluster of dieffenbachia sprouted from the forest floor. I was not alone – the dieffenbachia assured me someone had been here before. I had been here before. As I was now.
Beyond the dieffenbachia I stumbled across the coral paver walkway. Literally. The walkway was obscured under leaf litter and ground growth, the edge still able to catch my foot. Filtered sunlight dappled the sea of thelypteris ferns around me. A Micronesian dove cooed in the distance.
I followed the trace of the coral pavers up the ridgeline, as I had some 30 years earlier. Where the pavers ended, however, the tree was not on my left as my memory had expected. I looked around, but could not see the tree. Turning right, I headed further up the slope. | The tree was still there. I scrambled up behind Walung Elementary School, following tracks that led to a bamboo grove. From there, I made my way to the ridgeback. The shadowy figure of a dark concrete water catchment tank could be made out.
Just past the tank, a cluster of dieffenbachia sprouted from the forest floor. I was not alone—the dieffenbachia assured me someone had been here before, as had I.
Beyond the dieffenbachia, I literally stumbled across the coral paver walkway. The walkway was obscured under leaf litter and ground growth, the edge still able to catch my foot. Filtered sunlight dappled the sea of thelypteris ferns around me. A Micronesian dove cooed in the distance.
I followed the trace of the coral pavers up the ridgeline, as I had some 30 years earlier. Where the pavers ended, however, the tree was not on my left as I had remembered. I looked around, but could not see it. Turning right, I headed further up the slope. |
Gemini performed nicely at tightening up the prose, changing a hyphen to an m-dash, while leaving the unrecognized word thelypteris alone (the spell checker flags that word as an error). Is this still my writing? Is writing improved by an editor still the original author's writing? Spell check, grammar, check, and now editorial feedback. These are now a part of word processing software, whether Microsoft Word or Google Docs.
The next step beyond an editor is the laboratory report shared above where the large language model writes the document based on requirements input by the user. This is early days in large language models, these models will only get better and more integrated into authoring technologies. The large language models are being integrated with reasoning engines to reduce the known issue of large language model hallucinations. Students will live in a world with these technologies. Those who know how to use these tools will outcompete those who do not.
Can and should students use spell checkers and grammar recommendations? Can and should students use editors to improve their writing? Can and should students use large language models to produce writing? Where on this technological spectrum does one say, "This, and no further." And why there? Not to mention questions as to how to enforce that decision. The students are already aware of these tools.
In the physical science class the students were asked to write their own reports using their own words. Enforcement is nothing more than an honor system. The students were told that if they choose to use an editor, they should write their report first and then see what the editor suggests.
Comments
Post a Comment