BI's use of AI -- What We Are Doing and Why

The Problem:

When Guy and Heidi retired from the University of Colorado in 2019, we lost our non-profit status and thus our ability to raise funds to support graduate students. So everything that has been done on BI since that time has been done by us, or by people who allowed us to post or repost their essays and interviews. Although that has been very rewarding, we also ended up leaving many projects undone, including the Guide to Constructive Conflict. Truth be known, even with nonprofit status and graduate research assistants, we almost always had difficulty raising sufficient funds, given the chronic underfunding that has always characterized the conflict resolution, peacebuilding, and democracy building fields — at least after the Hewlett Foundation dropped that funding focus. (BI was originally funded by a series of very generous grants from the William and Flora Hewlett Foundation, until they ended their Conflict Resolution Funding Program. 

Our Solution:  

Ever since BI's beginning in 2001, we have been trying to use BI to help students, scholars, practitioners, and indeed, the general public, learn what the conflict and peace fields know about more constructive ways of dealing with difficult and intractable conflicts. As societal polarization has intensified and social cohesion has broken down, helping people learn and utilize what we collectively know about more constructive approaches to conflict has become critically important.  AI has given us a powerful new tool to do this. It not only enables us to work around existing funding constraints; it is also enabling us to do things that would have been impossible to do before, even with generous funding. 

When I (Heidi) first started experimenting with AI a few months ago, asking it to write introductory essays for the Guide, it didn't do well.  I decided, though it was slow going, that I was going to have to write those essays myself. AI, despite the hype, simply wasn't up to the task.

But then I tried again in June of 2026, and was amazed.  The essays ChatGPT is now writing are excellent: they pull in most or all of the key ideas I would have included, based on material in BI and other credible sources (often including things I likely would have missed. The essays are reasonably well written, and neither the content nor the sources are bogus (products of what are called "hallucinations," of which AI is  infamous). (We learned on one of Yasha Mounk's podcasts that hallucinations are largely a thing of the past, and so far, with our testing, that seems to be right.)

So, suddenly, we have a way of overcoming our lack of funding and have what amounts to an army of assistants to help us flesh out various parts of BI. As you'll see below, we are also using AI to better organize BI (making its many resources more findable) and hope to use it to write spin-off articles, aimed at specific audiences, such as our generally accessible Things YOU Can Do to Help articles, which are aimed at the general public.

Our Guidelines and Guardrails

We are trying to be very careful in our use of AI. We give it a lot of background in our prompts, so it knows what we are looking for and what we consider to be credible sources. We always ask it to show its sources, which we check. While we found hallucinations three or four months ago, we haven't found any in June 2026. But we will continue to check every source.  We also carefully read every essay, edit them as necessary (not much has been necessary), and we often add in a little of our own writing, sometimes a lot.  But we are always public about how much of each article is written by AI.  At the bottom of each AI article, we show the prompt that was used to generate the article.

Other Ways We Are Using AI

Our first extensive use of AI came when Guy updated the entire site from Drupal 7 to Drupal 10.  Drupal is the content management system — a computer program — in which Bi is built.  Drupal 7 was being discontinued in January of 2025, so we needed to upgrade the system.  Being without funding, we couldn't hire a programmer to do this — Guy had to do it himself. And he is not a programmer, though he did earlier build the Drupal 7 system, so he knew something about how to do this.  But the transition did not go easily, as the directions were unclear and things that were supposed to work did not.  He turned to ChatGPT like he might have turned to a skilled Drupal programmer, asking it seemingly endless questions about how to do things, what might be going wrong, and how to fix it.  It still took months, but we agree — if it hadn't been for AI, we probably would not have succeeded in making the transition to Drupal 10 at all.

A second successful use of AI was to run simulated "Common Ground Exercises." Here we gave Claude the detailed directions for our common ground exercise, and then asked it to compile, based on all that it has read, a report on how it thinks that a truly a representative group of Americans might answer our six questions. We first asked it to do that for climate change, then immigration, then DEI programs. Its original answer was 32 pages long. Newsletter  444 summarizes those results, and for those who are interested, Claude's full response for all three topics is posted here:  Claude Does The Burgess’s Common Ground Exercise on Climate, DEI, and Immigration.

Now Guy is using AI to better code the thousands of pages of content we have on BI, so we can make it easier to find what you need.  We are working to improve the categorization of our newsletters, our colleague and news/opinion posts, and all the "legacy" BI content — essays, case studies, book and article summaries, practitioner profiles, interviews, etc. This will enable us to greatly improve our "browsing system," as well as improving our search system substantially. We will post updates about those efforts as soon as they become available. 

Heidi is using AI to finish up writing the introductory essays for the Guide; and Claude had already suggested about 30 Guide topics that might be turned into new Things YOU Can Do to Help Pages (and we will likely ask it to create drafts of those pages for us). We will post updates about these project here too.

We haven't even figured out what else Claude, ChatGPT and other AI systems might be able to do for us. But we do know, with its help, we can make BI much more complete and much more easily navigable, and more user friendly than it ever has been.

The Downsides of AI

We do not deny or ignore the fact that AI has downsides.  It is putting people out of work, and it may put many more people out of work, although some things we have read recently suggest that threat may be overplayed. (See, for example, The Budget Lab report, Luis Garicana with Yasha Mounk discussing the economics of AI, and MIT's Technology Review report entitled "A reality check on the AI jobs hysteria."

AI certainly uses up a lot of resources — particularly energy and water.  But it turns out that all U.S. data centers combined used 260 billion gallons of water last year (a lot, yes, BUT) California almond farms used 1,300 billion gallons! So maybe we should stop eating almonds before we cut off data centers?  

Seriously, we know there are downsides to AI and legitimate reasons to be concerned about it. But it is here, and we are not going back, any more than we are going back on social media, cell phones, TV, cars, or airplanes. So, as we see it, we should learn to use it wisely, learn how to prevent its misuse by ourselves and others, and see how it can be utilized to improve our world. It's potential to improve BI far beyond anything we could do ourselves, is enormous!