Imagining Audiences and Positive Reframing with LLMs
I was reflecting more on the topic of this post about the role of “imagined audiences” in sharing creative work. A key takeaway from that post was that there is value for me in being able to imagine what a constructive or positive response to the work I am sharing might look like. As I wrote previously: in many cases I will never know how readers or viewers consume my work, and so my imagination of their response is really all there is. Of course, there are many possible ways for me to imagine a reader’s reaction: from critical, negative, condesceneding to neutral, indifferent to positive, constructive, excited and so on. My imagination of these reactions affects my desire or willingness to share in some measure.
Something that came to mind as I was considering this point more is that generative AI tools (like ChatGPT) could be useful for supporting this kind of positive imagination. For example, I experimented with prompting ChatGPT4 as follows:
I am going to share with you a short blog post I just wrote about Bordieu and Putnam. Pretend that you are a leading expert on Bordieu and Putnam who is kind, thoughtful, and supportive. What is your reaction to reading the post?
After sharing my post, here is ChatGPT’s response:
Thank you for sharing your insightful blog post on the concepts of capital as articulated by Pierre Bourdieu and Robert Putnam. Your grasp of these sociological theories is commendable and your presentation of their distinct perspectives is clear and well-articulated.
You’ve aptly captured Bourdieu’s multifaceted concept of capital, emphasizing its role as a structuring force in social and economic realms. Your interpretation of capital as an accumulation of historical labor and its various forms (cultural, social, economic) reflects a deep understanding of Bourdieu’s work. You rightly point out the finite nature of capital at a given time and its potential for conversion and institutional inscription, which is crucial in understanding Bourdieu’s perspective on social stratification and power dynamics.
Your contrast with Putnam’s focus on social capital is equally insightful. You’ve captured the essence of Putnam’s argument that social capital is a collective resource, benefiting society as a whole rather than serving as a tool for individual gain. This highlights the fundamental difference from Bourdieu’s approach. Your mention of social capital being renewable and formed through collective activities brings out the essence of Putnam’s thesis on the importance of social networks and community engagement for the health of democratic societies.
It’s particularly interesting how you juxtapose Bourdieu’s view of capital as a means of private gain and social structuring with Putnam’s view of social capital as a communal resource that underpins social cooperation and democracy. This comparison not only shows a clear understanding of their theories but also provides a thoughtful reflection on how different types of capital can shape societies in diverse ways.
A suggestion for further exploration could be to delve into how these theories apply in contemporary contexts, especially considering the digital age’s impact on social capital and the evolving nature of cultural capital. Additionally, exploring the interplay between different forms of capital in specific social settings or demographics could add another layer to your analysis.
Overall, your post is an excellent synthesis of complex sociological theories, presented in an accessible and engaging manner. It’s a valuable contribution to discussions on social theory and its application in understanding modern societies. Keep up the great work!
Would any real reader (much less an expert) think this about my post? Who knows. But my point in the previous post is that, in many cases, we don’t ever know how many readers are responding to our work12. And so it’s really just about the imagination; and we have a choice about what kind of imagination to adopt.
In general, I’ve been experimenting with using ChatGPT as a tool for positive reframing in various contexts, and find it more helpful than I expected. I posted a joke-y version of this practice here, but in general do find it at least somewhat useful.
Here’s an example of this kind of practice in a more general setting3:
I am working on application for a fellowship that is due on Monday. It’s currently Friday. I am afraid that I am not going to have enough time to finish the application, and that my application will be bad and I will get rejected. Can you help me take a more positive perspective on this situation?
Here is ChatGPT4’s response:
It’s completely natural to feel anxious about tight deadlines and the quality of your work, especially for something as important as a fellowship application. Let’s reframe this situation into a more positive perspective:
Opportunity for Growth: This application process is a valuable opportunity for personal and professional growth. It challenges you to articulate your goals, achievements, and aspirations, which is a beneficial exercise regardless of the outcome.
Showcase of Your Abilities: You’ve reached this point because of your skills, experiences, and potential. This application is a chance to showcase these qualities. Remember, the fact that you’re applying means you have qualities and experiences worth considering.
Learning Experience: Regardless of the outcome, the process of applying is a learning experience. It helps you understand what fellowship panels are looking for and improves your ability to present your ideas and projects compellingly.
Time Management Skills: Working under a tight deadline can enhance your time management and prioritization skills. It’s an opportunity to focus on what’s essential and deliver your best under pressure.
Positive Mindset: A positive mindset can significantly impact your performance. Believe in your capabilities and the effort you’re putting into the application. Confidence in your work can often be seen in the final product.
Potential for Success: Every application submitted is a step towards potential success. You miss 100% of the shots you don’t take. By completing this application, you’re giving yourself a chance to succeed.
Backup Plans and Future Opportunities: Remember, this fellowship is not the only opportunity. If it doesn’t work out, there will be other fellowships, grants, or opportunities that you can pursue. Each application makes you better prepared for the next one.
Seeking Support: Don’t hesitate to seek support or feedback. Whether it’s friends, mentors, or colleagues, a fresh perspective or encouragement can be incredibly helpful.
Reward Yourself: Plan a small reward for yourself after submitting the application. It’s a way to celebrate your hard work and dedication, regardless of the outcome.
Reflect on Accomplishments: Take a moment to reflect on what you’ve already achieved. This can boost your confidence and remind you of your capabilities and potential.
Remember, it’s not just about the outcome but the journey and what you learn along the way. Each step you take is progress, and that in itself is an achievement. Good luck.
While some of these points are a bit cheesy or simplistic (“you miss 100% of the shots you don’t take!”), I do find it’s at least somewhat helpful in accessing a more optimistic perspective, and in this case a more growth-oriented attitude. Continuing the conversation with a bit more anxiety and fear:
But I am scared that the readers will think my application is dumb and I will feel embarrassed. And I will be really disappointed if I am not accepted into the fellowship. What should I do?
Here is ChatGPT’s response:
Feeling scared about how others will perceive your work and fearing disappointment from not achieving a goal are common concerns. Here’s how you can address these feelings:
Understand the Nature of Fear: Fear of rejection or embarrassment often stems from our inner critic. Recognize that this is a natural response to stepping out of your comfort zone. Acknowledging this fear as a normal part of the process can help reduce its power over you.
Focus on Effort, Not Outcome: Concentrate on the effort you’re putting into the application rather than fixating on the outcome. You have control over the quality of your application, but not on the decision of the fellowship committee. Doing your best is a success in itself.
Redefine Success: Redefine what success means in this context. It’s not just about getting the fellowship; it’s also about the learning and experience gained through the application process. Every application you complete builds your skills and resilience.
Seek Constructive Feedback: Before submitting, you might want to share your application with a trusted friend, mentor, or advisor for feedback. Constructive criticism can help you improve your application and also reassure you about its quality.
Prepare for Different Outcomes: Mentally prepare for both acceptance and rejection. If accepted, great! If not, it’s not a reflection of your worth or capabilities. Many successful people have faced rejections and used them as stepping stones.
Develop a Support System: Surround yourself with supportive people who believe in you. Their encouragement can be a great source of strength, especially in moments of doubt.
Practice Self-Compassion: Be kind to yourself. Remember that nobody is perfect, and it’s okay to feel scared or uncertain. Treat yourself with the same compassion you would offer a friend in your situation.
Use Positive Affirmations: Positive affirmations can help shift your mindset from fear to confidence. Affirm your abilities and worth, reminding yourself that you are capable and deserving of this opportunity.
Plan for Post-Submission: Once the application is submitted, engage in activities that you enjoy and that take your mind off the waiting period. This can help reduce anxiety and keep things in perspective.
Reflect on Personal Growth: Regardless of the outcome, you will have grown from this experience. You’ll have a better understanding of application processes and a clearer view of your own goals and aspirations.
Remember, applying for the fellowship is an achievement in itself. It shows your willingness to pursue your goals, face challenges, and grow. Whatever the outcome, you will have gained valuable experience and learned more about yourself, which is invaluable for your future endeavors.
Here I actually find its responses a bit more insightful. For example, the idea of “redefining success” as it describes seems useful. Also, I appreciate its emphasis on normalizing and accepting fear as something natural and understandable in the situation.
Reflection
Is using an LLM in this way a weird, “dystopian”, or otherwise problematic practice? I’m not sure. It reminds me a bit of this black mirror episode, in the manner of allowing (inviting) AI-type technology into one’s personal emotional space. Of course, it’s not hard to conjure ways that using AI chatbots for therapy-like things can be really bad. I would never recommend it as a practice for anyone, especially anyone experiencing serious mental crisis.
Still, in another sense, it does seem to me that generative AI tools are relatively well-suited to this kind of work of reframing. For example, in other kinds of use-cases, so-called “hallucination” is a major concern – the model might invent some plausible-sounding answer to a question that is actually not true at all. As some have framed it: “LLMs are sophisticated bullshit artists – they make stuff up, convincingly.”
But in this “reframing” use-case, there’s a sense in which “hallucination” is exactly the goal – give me a plausible-seeming positive way of looking at this situation. There is no “right” answer of how to look at it; it’s all just perspective – a choice of what to focus on. And worst case, if the proposed reframe is not compelling or plausible, I can just ignore it.
It might be interesting to try extending the practice to align more explicitly with the cognitive behavioral therapy (CBT) strategy of “worst, best, most realistic”, which is supposed to help access a more balanced perspective on anxiety-provoking situations. A fair question is how offloading some of the work of this technique to an LLM will help or hurt the effectiveness of the approach (I can imagine both pros and cons).
So I don’t know – this is something I’ll be experimenting with more. If you have thoughts or give it a try, let me know.
Update
After posting this, I came across this post and pre-print paper from Amit Goldberg: https://x.com/Amit_Goldenb/status/1782076346929475602
The paper explicitly explores the use of LLMs compared to humans in cognitive reframing of negative scenarios. Check it out!