Generative AI Is Better At Form Than Content

Prudens Futuri prominently displayed in
the US Army War College's new academic building

All communications have at least two aspects, form and content.  Content is what we want to say.  It could be about politics, or philosophy, or here at the War College, rather predictably, about war.

Form is different.  The medium through which we express ideas impacts how we communicate those ideas. For instance, the same core concepts about politics, philosophy, or war can be conveyed very differently depending on whether they are presented through a 20 minute PowerPoint presentation or a 2,000 word paper. The form we choose defines how we communicate the message.

Let me give you an example.  My current favorite generative AI is Anthropic's Claude.  I asked it to explain the meaning of the US Army War College's motto, "Prudens Futuri," to the incoming class of students.  Here is what it came back with:

"The Latin phrase "prudens futuri" translates to "foreseeing the future" or "provident of the future." It refers to the wisdom and foresight required of effective leaders. Prudent leaders do not simply react to current circumstances - they have a vision for the future and proactively position their forces to capitalize on anticipated events. As Sun Tzu advised, foreknowledge enables leaders to seize opportunities and gain advantage. Prudens futuri requires gathering intelligence, conducting analysis, forecasting threats, and planning for contingencies. Leaders who excel at prudens futuri can execute bold, unconventional strategies to defeat enemies who rely on reactionary thinking. Prudens futuri epitomizes the innovative, forward-looking mindset the Army must adopt to maintain strategic dominance."

Not bad.  I also asked it to explain the same thing to a 13 year old:

"'Prudens futuri' is a Latin phrase that means "foreseeing the future." It's about being wise enough to predict what might happen next. Great leaders don't just respond to what's in front of them - they think ahead. Imagine you're a general planning to attack an enemy. A wise general would send scouts to spy first and learn the enemy's weaknesses. Then the general can make a smart battle plan to sneak around and surprise them from behind. The Latin phrase teaches that real leaders don't just react - they use their brains to see the future and stay one step ahead. That's what "prudens futuri," or "foreseeing the future," is all about!"

Also not bad--particularly given the change in audience.  

It is even possible to ask the same of a text-to-image AI like Midjourney, used to create the image below from the same prompt:

In this image, I am particularly taken with the way the AI combined historical references to foresight, such as the Oracle of Delphi, with more futuristic concepts like the (apparent) space elevator in the background.  In fact, I find asking text-to-image generators to imagine some futuristic scene to be a very practical way to fill in gaps in futures thinking.  I inevitably see something that makes sense in the image that was not accounted for in the text (kind of like The Spandrels of San Marco and the Panglossian Paradigm).

My point?  I have yet to have a generative AI create something that did not need tweaking, at the least, on the content side.  I have been really happy, however, with generative AI's ability to master particular forms.  

This is one of the reasons, I think, I have quite recently become a bit uncomfortable with policies that talk about citing a generative AI as if it were a source.  It is, I suppose...but it seems less of a source than Wikipedia, and, while I love Wikipedia and believe it is one of the great wonders of the modern world, I would not cite Wikipedia for anything other than background.  I require my students, for example, to find a reputable source to validate anything that a generative AI might come up with when making an estimate.  And, if you are going to make a student find a reputable source anyway, why would they need the generative AI at all?  The answer, of course, is for the form.  

This may not be true forever.  Generative AI is getting better at a brisk pace.  There may come a day when generative AI is looked upon as an authority, equal to peer-reviewed papers.  Until that time, we should still appreciate its talents for helping to craft the message. For now, generative AI is an unparalleled writing partner, not an independent thinker. By acknowledging its current limits alongside its awesome potential, we grant generative AI its proper place: revolutionizing how we communicate knowledge, while established methods still reign over what we know.

Why Good Data Isn't Enough (British Medical Journal And The University Of Michigan)

You are briefing the boss today and you are pretty excited.  You were tasked to take a hard look at two different ways of doing the same thing -- the "old way" and the "new way".  The old way was OK but your research clearly shows that the new way is much better.

You stand up in front of the boss.  You know you are speaking a little quickly (you may not even be pausing all that much) and your voice is probably a little higher than it usually is -- but none of that matters.  Your data is rock solid.

In fact, you have even put your great data into a pie graph that clearly identifies the validity of your position.  This is your ace in the hole because you know the boss loves pie graphs.

All of this explains why you are stunned when the boss decides to continue to do things the old way.

Two interesting studies, one quite old and one brand new, explain why what you said mattered far less than how you said it.

http://www.bmj.com/content/318/7197/1527.full

The first study, from 1999,"Influence of data display formats on physician investigators' decisions to stop clinical trials: prospective trial with repeated measure" from the British Medical Journal (hat tip to social network analysis expert Valdis Krebs and his prolific Twittering) asked a number of physicians to look at the exact same data using one of four different visualization techniques -- bar graph, pie graph, chart or "icons".  You can see the four different charts in the picture to the right.  Note:  The test subjects only saw one of these, not all four together at once.

Now, I admit, these charts are a little dense at first.  Basically you have 2 different groups, those who started the study with a good prognosis and those who started the study with a poor prognosis.  You also have those who received the old treatment and those that received the new treatment.

The question was, based on these results, do you continue this study or not?  The doctors involved in the study were all research physicians and used to seeing this kind of data and making these kinds of decisions. 

Despite the fact that the data was exactly the same in all four images and that the data was overwhelmingly in support of the new treatment option, there was a satistically significant difference in the accuracy rate of the physician's decisions based exclusively on how the data was presented.

The least accurate?  Pie and bar graphs.  Charts did OK but the best option was the "icons". 

This kind of iconic chart is probably new to many readers.  It shows the impact of the treatments on every single patient in the study.  While this kind of display yielded the most accurate results in the study, it was also the most disliked by the test subjects.

The overwhelming preference was for the chart, while a minority preferred the bar or pie graphs.  Not only did none of the participants indicate that they preferred the icons, a significant number of them expressed derision at the format in their after action comments.

This study reminds me of a series of studies conducted by Ulrich Hoffrage and Gerd Gigerenzer at the Max Planck Institute in Berlin that demonstrate that expressing statistics using "natural frequencies" (e.g. 2 out of 20 instead of the more common 10%) leads to better understanding and better (i.e. more "Bayesian") reasoning (Jen Lee, Hema Deshmukh and I were able to replicate these results using a typical analytic problem so I believe that this effect is important in the context of intelligence as well).

The second piece of research is from the University of Michigan's Institute For Social Research and is still in pre-publication review.  In what appears to be a very cleverly designed study, researchers looked at 200 telephone interviewers (100 male and 100 female).

They found that interviewers who spoke moderately fast, with lower pitched voices (if male) and with 4 to 5 natural pauses per minute were the most effective at getting people to listen to them.

Combining the results of these studies, it is easy to imagine that the most powerful presentation would be one using icons combined with a proficient speaker.  The opposite (as demonstrated in the story that started this post) could reasonably be expected to perform less well -- even if the information were exactly the same.

As I have said before, like it or not, it is not enough to have good info, you have to be able to communicate it effectively as well.  The flip side of this coin is equally important for intelligence professionals -- we may well be hard-wired to be biased towards high quality forms of communication, even if the quality of the content is second rate.