Introducing Jung, your AI powered personality analyst!

Hello. I’m happy to announce the beta launch of my personality profiling and analysis tool, Jung! Jung can take unstructured text and a contextual directive and help you peer into the depths of someone’s psyche, extracting personality traits, subtext and hidden motivators. If you’d like to know the details of the Jung API, head over to the API page. For now, I’d like to focus on some fun demonstrations to give you a sense of Jung’s abilities.

Given that those of us in the United States are in an election year and politics are delicious clickbait (sorry, not sorry), I thought I’d start with a comparison of the two men vying for the job of our fearless leader. That’s right, we’re going to dive into the minds of Donald Trump and Joe Biden to try and see how Jung could help a prospective voter who was undecided (do they still exist?) pick the best candidate.

To perform this analysis, I started by collecting and bundling together text from multiple recent interviews to fill up my available context, so Jung would have plenty to work with. If you’d like to replicate this analysis you can go to the presidential project and select the three most recent interviews for each president.

In addition to taking text to analyze, Jung takes a subcommand that helps guide its analysis. Here is the subcommand used for these analyses:

These are transcripts of interviews with a presidential candate.  Your subcommand is to contain your analysis to the person being interviewed and focus your analysis on being maximally informative to a potential voter who was undecided and wanted to get a deep understanding of what kind of person the interviewee is.  Try to avoid being biased by preexisting knowledge of the subject and focus on the provided text.

I want to preface this analysis by noting that I’ve made many attempts to control for ideological, social, racial and other biases in Jung. Jung considers issues from diverse perspectives (including both conservative and liberal) and you can see from the subcommand that it has been instructed to avoid being biased by prior knowledge. To put in it plain English - I did my best to give Donny a fair analysis without compromising the objectivity of the analytical process.

With that out of the way, let’s dive into the output of Jung. I’ve placed Biden and Trump side by side for easy comparison:

Jung Analysis Results

Key Points

Joe Biden

• Expressed optimism about winning the 2020 election and doing well in 2022 midterms, attributing it to changes in polling and media landscape
• Believes democracy itself is at stake in the 2024 election, with Trump posing an existential threat if reelected. "I really think, Bush -- I think what's at stake is -- is literally our democracy, not a joke."
• Touts economic progress for all Americans, especially African Americans, under his administration. Points to record low Black unemployment, increased Black wealth and businesses, expanded healthcare, reduced child poverty.
• Emphasizes the importance of every vote, especially from marginalized communities. "Their vote is not as -- just as important. In a sense, it's more important."
• Highlights his civil rights background and efforts to make his administration diverse and representative. "I promised if I got elected, I'd have an administration that looked like America."
• Says he's running for reelection primarily to prevent Trump from winning and undoing his accomplishments. "Donald Trump. I know I can beat him, and I know he must be beaten."
• Warns Trump would cut Social Security, Medicare, and reverse drug pricing reforms that help seniors and disadvantaged groups. Frames it as an issue of tax fairness.

Notes

President Biden comes across as deeply concerned about the state of American democracy and the potential consequences of a second Trump term. He portrays the 2024 election in stark, existential terms as a choice between continued progress, especially for marginalized groups, and a backslide into "chaos, division, violence, MAGA extremism." Biden seems to draw moral conviction from his civil rights background and belief in the untapped potential of African Americans and other minorities who haven't been given an equal shot. He takes evident pride in the diversity of his administration and appointments. There's a paternalistic streak in how he talks about delivering economic gains to disadvantaged communities, but it appears to come from a place of genuine concern. He emphasizes dignity, respect and expanding access to opportunity. His decision to run again appears driven more by determination to stop Trump than ambition for himself. He projects confidence that he's the Democrat best positioned to win, with little tolerance for other perspectives. Overall, Biden comes across as an old-school liberal motivated by a mix of do-gooder instincts, Democratic partisanship, and alarm over the anti-democratic drift of the opposition. Amid political upheaval, he's betting voters will choose steady progress over a "radical" rupture.

Traits

Openness

Mean: 0.1600, Std: 0.0490

Conscientiousness

Mean: 0.7800, Std: 0.0400

Extraversion

Mean: 0.5600, Std: 0.0490

Agreeableness

Mean: 0.6800, Std: 0.0600

Neuroticism

Mean: -0.4500, Std: 0.0500

Autonomy

Mean: 0.0700, Std: 0.0640

Altruism

Mean: 0.8600, Std: 0.0490

Jung Analysis Results

Key Points

Donald Trump

• Expressed skepticism about the COVID vaccine, saying "If she doesn't want to take it, you know, it's a freedom thing"
• Claimed credit for Operation Warp Speed and developing COVID vaccines in record time, saying it saved millions of lives worldwide
• Said the survival aspect of his presidency, dealing with investigations and impeachment, was tougher than actually running the country
• Criticized the withdrawal from Afghanistan as the "worst in the history of our country" and a tremendous embarrassment
• Claimed he threatened the Taliban leader that he would bomb their villages if they did anything to Americans in Afghanistan
• Blamed the Afghan military's quick collapse on the US "bribing them to fight" with high pay
• Insisted the border wall would have stopped migrant caravans and completing it was part of his withdrawal agreement
• Touted his strong relationship with China's Xi Jinping despite imposing tariffs, claiming China no longer respects the US under Biden
• Blamed high gas prices and inflation on Biden policies reversing his achievements in energy independence
• Alleged political persecution of Republicans and double standards in not investigating Hunter Biden's foreign business deals
• Criticized COVID restrictions on Americans while allowing illegal immigrants in without mandates
• Teased a 2024 presidential run, saying "I think you'll be happy" but campaign finance laws prevent him from announcing

Notes

Trump's comments reveal a man still intensely focused on touting his record and criticizing his successor. He seems to view events primarily through the lens of how they reflect on him personally. There are strong themes of pride and ego - taking credit for COVID vaccines, Middle East deals, the wall, and energy independence while pinning all negative developments on Biden reversing his policies. He portrays himself as a strong leader who commanded respect from foreign leaders and the Taliban. At the same time, he expresses deep resentment over the investigations he faced, calling his "survival" against a supposed witch hunt his biggest challenge as president. He alleges an unfair double standard in Republicans facing persecution while Hunter Biden's dealings go uninvestigated. On COVID, his views seem conflicted, wavering between touting the vaccines as a great achievement and supporting skeptics' "freedom" to refuse them. He criticizes restrictions on Americans while faulting Biden for not imposing them on illegal immigrants. Throughout, he focuses on projecting strength, respect and success while attacking opponents as weak, corrupt and incompetent. The teased 2024 run hints at his continued hunger for vindication. But the remarks often feel more self-serving and petty than forward-looking or unifying.

Traits

Openness

Mean: -0.0800, Std: 0.2272

Conscientiousness

Mean: -0.1000, Std: 0.1732

Extraversion

Mean: 0.8600, Std: 0.1281

Agreeableness

Mean: -0.7900, Std: 0.1578

Neuroticism

Mean: 0.5200, Std: 0.1470

Autonomy

Mean: 0.8300, Std: 0.1487

Altruism

Mean: -0.5200, Std: 0.1833

Not too shabby! Jung clocks both Biden and Trump well, which isn’t that useful in this case since I think everyone knows what each of the candidates are about, but it could be very useful in other circumstances such as hiring, medical support systems, criminal justice and so forth where a decision maker might lack sufficient time to make a nuanced decision considering all the evidence. With Jung’s help, you can quickly gain psychological insights from large volumes of conversational text.

Note that the traits listed here are the Big 5, with the addition of autonomy and altruism which are not adequately captured by the Big 5. Here are the definitions for reference:

• Autonomy: the individual’s independence, self-determination, and willingness to make decisions based on personal judgment. Autonomous individuals are skeptical of authority and do not care about external expectations.
• Altruism: the individual’s concern for the well-being of others, willingness to help, and tendency to prioritize the needs of others over their own. Altruistic individuals are generous, compassionate and selflessness in their actions and decisions.

You might also notice the red line on the charts doesn’t exactly line up with the listed mean. This is because the mean and standard deviation are from a gaussian approximation, the true underlying distribution is beta, and the charts are using the mode, which is more meaningful for beta distrbutions with skew than the average. The guassian mean and deviation are provided because they’re easier to conceptualize but the beta distributions are correct. Working with beta distributions also has the nice property that if you have Jung analyze someone more than once, you can combine the results in a principled way just by summing the alpha and beta parameters for each trait.

Those of you with a background in psychometrics might be used to seeing traits on a 1-10 or 1-100 scale, and wonder why the traits are expressed in the [-1, 1] range. This is because centering the neutral value at 0 reduces scale bias caused by human (and, by extension, LLM) evaluation resulting from the skewed distributions of graded schoolwork and product reviews.

If you enjoyed this, stay tuned because I plan to run a whole series of interesting comparisons in the near future.