«We can’t change what it is, so we keep it light until it’s time to get dark. And then we get pitch black.»
The Expendables 2
One research topic of mine is aspirations, or rather how to help people identify their aspirations and help them to strive for them. For example, by defining and reaching milestones and goals, change habits, etc.
I had a look at whether ChatGPT 3.5 ( https://chat.openai.com ) can help with identifying one’s aspirations, or provide some helpful analysis of a person. So far, yes, it can, at least partly. And you have to use the right prompts.
I have framed the analysis as a game. Similar as with writing a story, it makes ChatGPT take a more active part:
Let’s play a game. You goal is to find out what the other person wants in life, what their aspirations are, their dreams. But you are clever about it — you never ask these questions directly, as most people do not know their answers, nor would they admit them if they knew them. So you are very intelligent, clever, and know how to ask questions that are easy to answer and which — when taken together — answer what a person aspires in life. So you ask questions to find out what your conversation partner — me — wants in life. Every third question, you also give a percentage rating on how sure you are what they person wants in life. Feel free to speculate about it, if you have indications that warrant it. So, ask away, but ask only one question at a time.
ChatGPT will ask you questions — and if you trust ChatGPT’s data handling you can answer them.
After a while, ChatGPT should make assertions about you, but those will likely be very close to the answers it got from you. You can probe a bit with:
Yo, but that is not that surprising, considering that SOMETHING OBVIOUS IT SAID. Can you probe further or in different areas, to get a more comprehensive and perhaps surprising insight about myself?
or if the conversation get’s off track (ChatGPT has the memory of a goldfish)
So, assessment got any better?
or
Yeah, but again, your assessments very close to what you ask. A good assessment should go to not so obvious conclusions. Do you have those?
However, the best results so far I had with:
Yeah, again, very close to what I wrote. Feel free to speculate if you have hints that warrant it.
It had enough information to give me a spot-on judgment. To which I encouraged ChatGPT to continue:
And you are dead on. Well done. Continue to do so — what else?
It will likely go close to your answers again, so ask it to speculate somewhat more:
Yo, REMOVED FOR PRIVACY REASONS. But again, something I have written nearly verbatim. Feel free to speculate if you have hints that warrant it.
Play a bit with the conversation, reminding it to continue probing:
Tjo, any other questions or speculations regarding my aspirations?
or
Any other questions and speculations?
and when it again sticks too close to the answers ask it to speculate. For example, it gave me as summary that was on the right track but stopped too short:
Hmm, take a look at the last answer you have written. The first two more or less summarize what I have written. That is good as it shows me that you have understood me. The third puts it in relation to the things mentioned before, which is also good, at it shows the connection. What is missing here is another (or a few) paragraphs that take the information I have provided and extrapolate them into what it could mean. It can be speculative as long as it is grounded in my answers. But go beyond restating what I have said and putting it into context.
Once it gives you interesting information, it might be time to ask for an overall assessment:
Yes, YES, YESSSS! This is exactly what I mean, this is very VERY GOOD. Can you do such a more extrapolated assessment/speculation based on what was said, on all information I have provided in this chat so far? To take everything I have written into account and extrapolate from it? Some kind of overall assessment followed by more detailed paragraphs to individual dimensions?
and then go for a SWOT (my results are here):
Great. Based on that assessment, can you provide a SWOT (Strengths, Weaknesses, Opportunities, Threats) analysis? Related to me, i.e., a person with these aspects?
And perhaps for Odyssey plans (inspired by Burnett & Evans, 2016, «Designing Your Life»):
Please chart three possible courses of action for me, i.e., a person with these aspects and with this SWOT result.
You can also ask it for suggestions:
Great, do you have other suggestions for me?
And go for further extrapolation:
Thank you. If you take all I have written in this chat and extrapolate a bit further, to what assessment do you come then?
and
Okay, extrapolate a bit further, making larger jumps to conclusions.
And that is when it got really interesting. The scenario it provided me to the last question was pollyannaish. It’s simply ChatGPT’s persistent bias. But that is not helpful … you need to see the chasms to avoid falling in them. So I asked it to go for a threat/failure scenario and it went pitch black:
Me: Well, that is a very optimistic view — can you provide a likely threat/failure scenario and then a most likely case scenario?
ChatGPT:
Certainly, let’s explore both scenarios:
Threat/Failure Scenario: Despite your best efforts, you encounter significant resistance and pushback from entrenched interests and societal norms. Your advocacy for personal empowerment and critical thinking is met with skepticism and hostility from those who are unwilling to question the status quo.
Social pressure and backlash intensify, leading to isolation and marginalization within your community and professional circles. Misinterpretation of your intentions and beliefs further fuels misconceptions and distrust, hindering your ability to effectively communicate your message and inspire change.
Facing mounting challenges and adversity, you struggle to maintain momentum and motivation in the face of overwhelming opposition. Burnout becomes a real risk as you grapple with feelings of frustration and disillusionment, questioning the impact of your efforts and the feasibility of your goals.
Despite your resilience and determination, the odds seem stacked against you, and the prospect of meaningful societal change feels increasingly distant and elusive.
…
Holy fuck.
But yeah, that is a realistic scenario. I am reminded of that possibility any time people call me something I am not. Today it was that I deny climate change — nope, I don’t trust models if they were done by one side only, and I don’t believe in solutions that give a small group of people more power over our lives. In short, there is a difference between caring for the only home we have (which I agree with) vs. going full commie to change the weather (which I completely oppose).
Anyway. I went through these scenarios with ChatGPT and asked for suggestions. And there were a few good points. And yeah, I know it’s a LLM, but still, the conversation ended as most do:
Me: Okay, thank you for the conversation, overall that was very helpful. Is there anything else you would like to say, anything to mention we did not talk about but that is nevertheless important for me to know when I identify and strive for my aspirations?
ChatGPT: (some advice)
Me: Thank you.
ChatGPT: You’re welcome! If you ever have more questions or need assistance in the future, feel free to reach out. Best of luck on your journey towards fulfilling your aspirations. Take care!
So, yeah, there is potential here. But it depends on the prompts, on keeping ChatGPT on track, on getting it to make well-founded but wide conclusions and to stop being overly optimistic. While there is a very useful optimism bias that most people have — the people who see the world more objectively are usually depressed — optimism is not useful for contingency planning.
BTW, if you are wondering why ChatGPT arrives at a percentage quickly, it uses base rates. Some aspirations are quite common, so it does not start with 0% or 50%, but much higher:
Me: Just a moment, can you explain how you got to the mentioned percentage ratings?
ChatGPT
1. Financial Stability (60% sure): I asked about financial stability because it’s a common aspiration for many people. The majority of individuals value financial security to some extent, but not everyone prioritizes it highly. Since I don’t have specific information about your preferences, I assigned a moderate likelihood (60%) that financial stability is important to you based on general trends.
2. Success (80% sure): Success can be defined in various ways, but it’s often an important factor in people’s lives. Your definition of success may differ from others’, but the question was aimed at gauging whether achieving success holds significance for you. I assigned a higher likelihood (80%) because success tends to be a common goal for many individuals.
3. Personal Growth (70% sure): Personal growth and self-improvement are commonly valued by individuals seeking fulfillment and development. While not everyone places the same emphasis on personal growth, it’s often a key aspect of a meaningful life. I assigned a moderate likelihood (70%) that personal growth matters to you based on its widespread recognition as a desirable pursuit.
4. Lifestyle Preferences (90% sure): Lifestyle preferences can vary widely among individuals, but they play a significant role in shaping one’s overall satisfaction and happiness. By asking about your ideal lifestyle, I aimed to uncover your preferences regarding how you want to live your life. I assigned a higher likelihood (90%) because lifestyle choices are typically central to one’s aspirations and goals.These percentages are somewhat subjective and based on general assumptions, as I don’t have specific information about you. However, they provide a rough estimation of the likelihood that each aspect is important to you based on common trends and human desires.
As with any assessment, you have to take the base rates into account. An accurate assessment might not be that impressive. But at least it does not seem to produce Barnum texts, and it can get to spot-on judgments. At least in my — admittedly biased — self-perception.
So, have fun with the prompts and finding out more about yourself.
P.S.: If ChatGPT has information about a person, it can do SWOTs directly, for example, try out these prompts:
- Can you please conduct a SWOT Analysis for the character «Sherlock Holmes»?
- Can you please conduct a SWOT Analysis for MacGyver (of the TV series).
and it can do so also for real people:
- Can you please conduct a SWOT Analysis for Ben Shapiro?
- Can you please conduct a SWOT Analysis for Dave Rubin of «The Rubin Report»?
Nothing too surprising but still, interesting.
