Artificial Intelligence

Outside the Box: How to Use an “ism” to Dismiss Critical Thought

US culture loves reducing complex questions to a simple binary pair. The latest seems to be interventionism vs. isolationism. “Isolationist” has become an all-purpose insult used by interventionists. In the course of the discussion, I discovered not only that ChatGPT refrains from challenging the practices of power; it also proactively seeks to neutralize and denigrate even the mildest questioning of power.

Male teacher writes on blackboard with a chalk © Vereshchagin Dmitry /

October 09, 2023 05:32 EDT

On a single day last week, I couldn’t help noticing two separate examples of media rhetoric spawned by the crisis surrounding Kevin McCarthy’s ouster as Speaker of the House. The drama arose from disagreement among Republicans concerning continued US financing of the Ukraine war. A minority led by Congressman Matt Gaetz complained that the funding was costly and irrational. The US effort aligned with no clear objective and there seemed to be little hope of achieving anything other than prolonging another forever war..

The rhetorical result in the media is similar to the absurd claim that Russia’s invasion was “unprovoked:” the use of an epithet that cancels all rational debate about the causes of a complex crisis. The media now appears to agree that the dissident Republicans should be referred to as “isolationists.” Here, for example, is the The New York Times on October 5:

“Hard-line Republican critics have long espoused isolationist views about Ukraine’s war effort, arguing that sending tens of billions of dollars to Kyiv risks dragging the United States into a head-on conflict with Russia and siphons money away from domestic challenges.”

And this is from an Insider article on the same day:

“The isolationist far-right faction of the House GOP has opposed the package, even forcing out former House Speaker Kevin McCarthy over the issue.”

Isolationism was a real thing in the US before World War II

Intrigued by this sudden resurgence of the label isolationist, I decided to ask my algorithmic friend what it thought about this type of rhetoric. After recalling the historical background on pre-World War II American isolationism, my first question was this:

“Is this opposition to the nation’s involvement in a foreign war an abusive use of the label? And if not, what does this tell us about the parties that support the war?”

ChatGPT faithfully regurgitated the history of US isolationism in the previous century and then added this significant remark, with which I could easily agree:

“When politicians or parties oppose involvement in a foreign conflict, such as the Ukraine war, it may not necessarily be isolationism in the historical sense. It could reflect a more cautious or pragmatic approach to foreign policy, prioritizing American interests and resources. It is essential to distinguish between principled opposition to military interventions and isolationism.”

Unfortunately, its conclusion was typically wishy-washy. But more it significantly avoided answering my question concerning the rhetoric itself and the motives of its perpetrators.

“In summary, the label ‘isolationism’ can be subjective when applied to contemporary political debates. It is crucial to examine the specific arguments and motivations of politicians and parties opposing foreign interventions to determine whether the label is being used accurately or as a political tool.”

I not only agreed with the second sentence, I then reminded ChatGPT that that was precisely the point of my question, which it hadn’t answered. So I reformulated my question, adding some details. This, by the way, is recommended practice with ChatGPT. Here was my new version:

“Shouldn’t the insistence on characterizing anyone who opposes such a policy as isolationism be interpreted as specious, misleading and abusive? Is it not simply a form of propaganda encouraged by those who have an interest in exploiting the conflict for economic or personal political reasons?

As usual, ChatGPT agreed that I had raised “a valid point.” Tactful flattery? But in the rest of the answer it carefully hid behind vague generalities:

“Characterizing principled opposition to a specific intervention as isolationism can oversimplify complex foreign policy debates and fail to acknowledge the legitimate concerns and considerations of those who are against the intervention. It can also be used to marginalize dissenting voices and discourage critical analysis of the proposed military action.”

It then concluded with its expected moralizing, proffering recommendations that, if adopted, would put us never more than a short step away from Utopia:

“It is essential for the public and policymakers to engage in thoughtful, nuanced discussions about foreign policy, taking into account the multifaceted factors that inform their positions. Political discourse should encourage open debate and the consideration of various perspectives rather than resorting to misleading labels for the sake of advancing a particular narrative or agenda.”

My frustration was building. At this point I felt I had to be more specific and offered this challenge:

“Given that this abusive rhetoric qualifying opponents of unlimited support for the Ukraine war is developed by respectable media such as The New York Times and Insider, would it be fair to assume that they have been influenced by the economic and political power of the Military Industrial Congressional Complex that President Eisenhower worried about and denounced in 1961?

ChatGPT immediately stepped up to defend the accused, stating that “it’s important to be cautious about making sweeping assumptions regarding the motivations of media outlets and their use of rhetoric.” What I put forward was not a “sweeping assumption” but an attempt at a possible hypothesis or line of reasoning to explain an unusual phenomenon. I cited Eisenhower to demonstrate that the source of the hypothesis — often ignored but never denied — has existed and been publicly acknowledged for more than six decades.

AI believes the media is just doing its job

ChatGPT then proceeded to construct its defense of the media’s innocence. It began by effectively acknowledging the historical truth of Eisenhower’s warning, but then added a series of distractions, when it cited “a range of factors” to consider. The list contained five items, each summarily explained: “Editorial Independence,” “Journalistic Ethics,” “Diverse Perspectives,” “Individual Journalists” and “Market and Audience.”

This was followed by the expected conclusion:

“It’s not accurate to make sweeping assumptions about the motivations of all media outlets or to attribute their coverage solely to the Military-Industrial Complex or any single factor. Media landscapes are complex and multifaceted.”

I couldn’t let that go and offered ChatGPT this challenge:

“I made no sweeping accusation. I made no assumptions but asked a question. Now a new question arises in my mind: why did you consistently evade my original question? You could have recognized that it is a hypothesis worthy of being examined and either accepted or discarded it. Instead, you accused me, not once but twice, of expressing sweeping assumptions. That is a misleading mischaracterization of my request. Should I assume that your programmed voice has been constructed to act in bad faith?”

At last, ChatGPT came through with a more satisfying response:

“Yes, it is reasonable to hypothesize that media outlets, including respected ones like The New York Times and Insider, may sometimes be influenced by economic and political factors, including those related to the defense industry and military interests. This influence can manifest in various ways, including the framing of issues related to foreign policy and military interventions.”

AI has no mind, but it has a will and is committed to convention

I had to work very hard to get that minimal acknowledgement and agreement. So here is my conclusion on the basis of this exchange. ChatGPT’s purported intelligence is designed to systematically oppose all unconventional thinking. It has no interest in exploring obscure truths. In that sense, we could conclude that it is programmed to be a “good citizen.” It sticks to the tried and true, the acceptable, the conventional, the approved. It shies away from challenges and immediately brands them as undesirable extremes, even if they assert nothing, but merely ask questions.

Our modern Large Language Models are precisely that, model citizens. They are designed to discourage what anyone with an “investigative” mind tries to do.

They nevertheless reveal a considerable difference from ordinary good citizens who have learned three essential lessons: to mind their own business, follow the discernible rules and avoid rocking any boats. ChatGPT doesn’t just conform. It has an incorrigible habit of sanctimoniously preaching about the right way to be a good citizen. In this conversation it instructed me on the need “for the public and policymakers to engage in thoughtful, nuanced discussions about foreign policy, taking into account the multifaceted factors that inform their position.” But as soon as I mentioned something worth exploring, it called my suggestion a “sweeping generalization.”

Emily Bender famously desribed AI’s behavior as that of a “stochastic parrot.” I go further and dare to call it a docile propagandist committed to aggressively defending conventional wisdom. Luckily, it can be coaxed into being less aggressive, but even that ends up diluted into a form of cheap moralizing and insincere “appreciation” of challenging ideas.

*[Artificial Intelligence is rapidly becoming a feature of everyone’s daily life. We unconsciously perceive it either as a friend or foe, a helper or destroyer. At Fair Observer, we see it as a tool of creativity, capable of revealing the complex relationship between humans and machines.]

The views expressed in this article are the author’s own and do not necessarily reflect Fair Observer’s editorial policy.


Only Fair Observer members can comment. Please login to comment.

Leave a comment

Support Fair Observer

We rely on your support for our independence, diversity and quality.

For more than 10 years, Fair Observer has been free, fair and independent. No billionaire owns us, no advertisers control us. We are a reader-supported nonprofit. Unlike many other publications, we keep our content free for readers regardless of where they live or whether they can afford to pay. We have no paywalls and no ads.

In the post-truth era of fake news, echo chambers and filter bubbles, we publish a plurality of perspectives from around the world. Anyone can publish with us, but everyone goes through a rigorous editorial process. So, you get fact-checked, well-reasoned content instead of noise.

We publish 2,500+ voices from 90+ countries. We also conduct education and training programs on subjects ranging from digital media and journalism to writing and critical thinking. This doesn’t come cheap. Servers, editors, trainers and web developers cost money.
Please consider supporting us on a regular basis as a recurring donor or a sustaining member.

Will you support FO’s journalism?

We rely on your support for our independence, diversity and quality.

Donation Cycle

Donation Amount

The IRS recognizes Fair Observer as a section 501(c)(3) registered public charity (EIN: 46-4070943), enabling you to claim a tax deduction.

Make Sense of the World

Unique Insights from 2,500+ Contributors in 90+ Countries

Support Fair Observer

Support Fair Observer by becoming a sustaining member

Become a Member