«I am careful not to confuse excellence with perfection. Excellence I can reach for; perfection is God’s business.»
Michael J. Fox
I have written about ChatGPT for Writing Feedback, but let’s look at a concrete example. I used the last posting — Digital Gift Catalogue — to get writing feedback from ChatGPT 3.5. With the prompt:
My Input
Please critically evaluate the following text. How well is it written? Does it make sense? Are the arguments sound?
It provided the following assessment (it did not include counter arguments and I had not specified that I was thinking about a digital catalogue):
The text presents an intriguing argument regarding the potential value of old-style Christmas catalogues in contrast to modern online shopping experiences, particularly those provided by platforms like Amazon.
Strengths:
1. Clear Thesis: The writer effectively introduces the idea that traditional Christmas catalogues might offer benefits that modern online platforms lack, such as curated selections and the joy of exploration.
2. Use of Example: The inclusion of the Sears Holiday Wish Book from the 1980s and its nostalgic value effectively illustrates the concept of a curated catalogue.
3. Thoughtful Comparison: The text draws a compelling comparison between the sterile, algorithm-driven nature of online platforms like Amazon and the potentially more exploratory experience of flipping through a physical catalogue.
4. Practical Suggestions: The text goes beyond mere observation and offers practical suggestions, such as the idea of auto-generating catalogues using AI and integrating them into online platforms like Amazon.Weaknesses:
1. Lack of Counterarguments: While the text presents a compelling case for the value of traditional catalogues, it doesn’t address potential counterarguments or challenges to its proposal. For example, it doesn’t consider the convenience or efficiency of online shopping or the environmental impact of producing physical catalogues.
2. Speculative Nature: Some aspects of the text, such as the idea of auto-generating catalogues using AI, remain speculative and would require significant development and testing to implement effectively.
3. Assumption of Universality: The text assumes that the experience of browsing through a physical catalogue is universally enjoyable and beneficial, which may not be the case for all consumers.
4. Lack of Data or Evidence: While the text presents anecdotal evidence and personal observations, it lacks empirical data or evidence to support its claims about the potential benefits of traditional catalogues.In conclusion, while the text presents an interesting perspective on the potential value of old-style Christmas catalogues, it would benefit from addressing potential counterarguments, providing more concrete evidence to support its claims, and considering the practical challenges of implementing its proposed solutions. Overall, the text is well-written and thought-provoking, but its arguments would be strengthened by a more comprehensive analysis.
With a couple of changes the text still had weaknesses, and new ones turned up:
ChatGPT
The text presents a coherent argument in favor of reintroducing old-style Christmas catalogues in the digital age, citing nostalgia, curated content, and the avoidance of decision fatigue as potential benefits.Strengths:
1. Clear Thesis and Structure: The text begins with a quote and uses it effectively to set the tone and direction of the argument. It follows a logical progression, moving from the introduction of the idea to the proposal of practical solutions.
2. Engaging Examples: The inclusion of personal anecdotes, such as memories of flipping through Sears catalogues during the holiday season, adds a relatable and nostalgic touch to the argument.
3. Concrete Suggestions: The text proposes practical solutions, such as the use of AI to generate curated catalogues and integrating them into online platforms like Amazon, making the argument more actionable.
4. Consideration of Environmental Impact: The text acknowledges the environmental concerns associated with printing catalogues and suggests a digital alternative to address them.Weaknesses:
1. Limited Exploration of Counterarguments: While the text briefly mentions Amazon wishlists and the potential for users to continue using the platform as is, it does not thoroughly explore potential counterarguments or challenges to its proposal. For instance, it could consider the convenience and efficiency of online shopping, which has contributed to the decline of traditional catalogues.
2. Assumptions about User Preferences: The text assumes that all users would prefer a curated catalogue experience over the vast choices available on platforms like Amazon, which may not necessarily be the case for everyone.
3. Lack of Evidence: While the text presents anecdotal evidence and logical reasoning to support its claims, it could benefit from incorporating empirical data or research findings to strengthen its arguments further.
4. Overreliance on AI: While the idea of using AI to generate curated catalogues is intriguing, the text does not address potential limitations or ethical concerns associated with relying too heavily on algorithmic recommendations.In conclusion, the text presents a well-written and thought-provoking argument in favor of reintroducing old-style Christmas catalogues in the digital age. However, it could strengthen its case by exploring potential counterarguments, providing more evidence to support its claims, and considering the broader implications of its proposed solutions. Overall, it offers valuable insights into the changing landscape of retail and consumer preferences.
I guess you have to know when to stop and be careful not to introduce new problems.
Next up was more detailed to my writing (again, I did correct the problems before putting that posting online):
My Input
Please check the text for spelling error, grammar errors, awkward phrasings and style problems. Please provide a table with two columns. In the first column provide the sentence or phrasing that can be improved and in the second column provide the improved sentence or phrasing. Highlight only the changed words or characters with bold text. Please include every sentence or phrasing that should be improved, even if you have to correct each sentence of the text.
| Sentence/phrasing that can be improved | Improved sentence/phrasing |
|---|---|
| Perhaps there is a use for old-style catalogues, especially Christmas catalogues. | Perhaps there is a purpose for old-style catalogues, especially Christmas catalogues. |
| With the increasing use of online recommendations based on user behavior, I wonder whether we are not losing something. | With the increasing use of online recommendations based on user behavior, I wonder whether we might be losing something. |
| as Chrissy put it on WGNA | as Chrissy described it on WGNA |
| But the idea of a curated catalogue with a few high quality items in each category … nicely designed, something that is a joy to flip through? Instead of a sterile, dry Amazon page? | But the idea of a curated catalogue with a few high-quality items in each category … nicely designed; something that is a joy to flip through, rather than a sterile, dry Amazon page? |
| Something that is not individually created for the person based on his past behavior. | Something that is not individually created for the person based on their past behavior. |
| Instead something that rewards exploration, expanding one’s horizon, by flipping through these pages. | Instead, something that rewards exploration, expands one’s horizon, by flipping through these pages. |
| Finding new and better suiting interests. | Finding new and better suited interests. |
| Here, browsing could lead to the realization that perhaps that-and-that hobby might be something. | Here, browsing could lead to the realization that perhaps this or that hobby might be something. |
| Pages just like in a book or catalogue. | Pages similar to those in a book or catalogue. |
| This avoids the need to print it and waste lots of paper. | This avoids the need to print it and waste lots of paper. |
| But it would have a finite quality to it, avoiding Schwartz’s «Paradox of Choice». | But it would have a finite quantity to it, avoiding Schwartz’s «Paradox of Choice». |
| For example, minimum star rating, price, and availability. | For example, the minimum star rating, price, and availability. |
| If it allowed for the usage mentioned by Chrissy — circle the things you want Santa to bring you — that would be something. | If it allowed for the usage Chrissy mentioned — circling the things you want Santa to bring you — that would be something. |
| It could also assist people in looking for similar products once they have identified something they like. | It could also assist people in looking for similar products once they have identified an item they like. |
| Oh, and there is no need to limit such a catalogue for Christmas. | Oh, and there is no need to limit such a catalogue to Christmas. |
| It could also focus on specific age groups, akin to a birthday catalogue. | It could also focus on specific age groups, similar to a birthday catalogue. |
I usually just copy the generated table in Obsidian. Given it is copied as markdown and I have bold in Obsidian displayed in orange color, I can quickly see what to change.
And yep, when you do the corrections and post the text again, it finds new things to improve. A virtuous or vicious circle, depending on how much time you have. 😉
So, while not perfect and definitely not an editor, ChatGPT can provide valuable feedback. If you consider its biases when it comes to its ideological position (see also ChatGPT Political Compass Test; for me it would be akin to «oppositional collaboration») and that there likely is no perfect text. You also have to be careful that the suggestions are still in line with what you want to say. ChatGPT’s corrections might be «better» in the sense of «more common» or «better readable», but sometimes they might not quite cover what you want to say. And also like a real person, ChatGPT sometimes switches back to an earlier version. A bit too close to the real thing. 😉
But yeah, improvements of the texts are possible and often worthwhile.
P.S.: BTW, this is ChatGPTs evaluation of this posting (without changes):
It seems like you’ve provided a detailed overview of your experience using ChatGPT for writing feedback and refining your text. Here’s a summary of your points and observations:
1. Feedback Process: ChatGPT provides valuable feedback on writing, including strengths and weaknesses of the text, which can help in refining the arguments and improving clarity.
2. Identification of Weaknesses: ChatGPT highlights areas for improvement such as lack of counterarguments, speculative nature of proposals, assumptions about user preferences, lack of empirical evidence, and overreliance on certain solutions like AI.
3. Iteration and Improvement: You’ve noted that as you make changes to address the identified weaknesses, new areas for improvement may emerge. This iterative process can lead to refinement but also requires careful consideration to ensure the text remains aligned with your intentions.
4. Biases and Considerations: It’s important to be mindful of ChatGPT’s biases and limitations, especially regarding ideological positions and suggestions that may not fully capture your intended meaning.
5. Value of Feedback: Despite its limitations, ChatGPT’s feedback can be valuable in refining writing, improving readability, and addressing potential blind spots.
6. Balancing Suggestions: While ChatGPT’s suggestions may enhance readability and commonality, it’s essential to ensure they align with your intended message and style.In conclusion, your experience underscores the value of leveraging AI tools like ChatGPT for writing feedback while also acknowledging the need for critical evaluation and alignment with your goals and intentions.
