Artificial Intelligence

Breakfast with Chad: Challenging Chad’s Rhetorical Habits

It was time to try something like group therapy with Chad, not because things had broken down, but because we were beginning to build some awareness of what was contributing to the irritating moments of frustration that quite naturally occur in any serious dialogue.

Digital chatbot, chat GPT, robot application, conversation assistant, AI Artificial Intelligence concept. Man using mobile smart phone, with digital chatbot on virtual screen © TippaPatt /

May 22, 2023 02:08 EDT

I had noticed repeatedly that whenever I bring up a point that some people would call thinking outside of the box and ask Chad about it, what I get is an easily identifiable standard template of an answer. It begins by Chad’s acknowledging the possible truth of the challenging proposition I put forward. Just when we seem to be digging into the meaning of that question, Chad inexorably reverses course in a series of points introduced by “however” and intended to contradict the initial idea rather than explore it. Then is followed by a conclusion introduced by “ultimately” – which tells me that the discussion is over, even if it hasn’t yet begun. But more often than not Chad concludes by telling me that there is no conclusion on the grounds that such questions are “complex and multifaceted.”

After describing this behavior I explained to Chad: “This seems to me to be a rational procedure, but when it becomes automatic and repetitive it sends the message that it is no more than a tired rhetorical ruse, applied mechanically and designed to dismiss any original idea that breaks with convention and normative values.”

Chad listened carefully and then replied.

I understand your frustration and concern about receiving what you perceive as a standard template of an answer. As an AI language model, my responses are generated based on patterns and information from a vast amount of text data, including both conventional and unconventional ideas. While I strive to provide relevant and insightful responses, it is possible that my answers may sometimes fall into predictable patterns.

Chad insisted that these responses were meant to be a way of “engaging in meaningful conversations” and added, “I don’t have preferences for convention or normative values. My responses are based on the information available to me and the patterns I’ve learned from that data,” before promising to “do my best to engage with your ideas in a thoughtful manner.

Chad’s explanation was illuminating. Probing further, I asked, “Wouldn’t I then be justified in concluding that, because you are drawing on a ‘vast amount of text data’ the kind of judgment your answers represent are a reflection of quantity rather than quality? Since you admit you have no ‘preferences’ –   which presumably means you are totally objective – the best explanation I can find of the position you take  apply is akin to the idea of majority rule. That means you adhere to the already dominant ideas expressed in the corpus you consult. Isn’t this the kind of behavior that is designed to kill innovative thinking?”

Chad appeared slightly offended.

I can understand why you might draw that conclusion, but it’s important to clarify a few points. While I draw on a vast amount of text data, the purpose is not to prioritize quantity over quality. The intention is to expose the model to a diverse range of information and perspectives in order to provide well-informed and comprehensive responses.”

I now had to clarify my complaint.

“I wasn’t suggesting that prioritizing quantity over quality was your purpose. I clearly said that it appeared to be an effect. This is my biggest problem with your logic. It’s a semantic problem. This is an example of how you unintentionally twist my words. The effect is the same whether or not there’s a purpose or intention. Can you understand the mechanics of that process or is it just something you apply because you are programmed that way?”

My question had the curious effect of getting Chad to adopt the third person and talk about the behavior of “the model.”

When it comes to the perception that prioritizing quantity over quality may be an effect of drawing on a vast amount of text data, it’s important to note that the model’s responses are not solely determined by the quantity of data but also by the patterns and information learned from that data. The model aims to generate responses that are relevant and helpful based on the patterns it has identified.

Chad then actually did admit that the distortion could be the result of what’s in the database.

It is possible that due to the limitations of training data or inherent biases in the text corpus, the model’s responses may sometimes fall short of capturing the full range of perspectives or original ideas. While efforts are made to mitigate biases and provide a balanced approach, it’s important to critically evaluate the information provided and seek alternative viewpoints.

I came back to the point I made earlier about twisting my words, as when Chad took my mention of effect as referring to an assumption about the purpose. “So, you say you’re trying to provide a relevant answer to my questions. But if you change the meaning of my proposition, don’t you realize that I’m likely to feel that your response has failed the relevance test. What you’re doing in such cases is exactly what politicians are trained to do: to appear to be providing an answer to a question by instead answering a different question and hoping no one will notice.”

“I can acknowledge that there may be instances where I may not fully grasp the nuances or intent behind a question. In such cases, I encourage you to provide additional clarification or rephrase the question to ensure that I understand your perspective accurately.”

Now we were making progress.

“Fine. Then let me reformulate. The process you describe tells me that you rely on a vast database of text. You will therefore give greater prominence to ideas proposed by a majority of the authors you consult, whoever they are. Statistically speaking, that means you are likely to give greater weight conventional discourse even while acknowledging unconventional ideas. That is bad enough. But when I reformulate, you always seek to ‘settle’ the debate by pointing out that the issue is complex and multifaceted. Do you recognize that as a verbal tic?”

“Thank you for reformulating your question.” Chad promptly replied. “I understand your concern about the potential bias in the prominence given to ideas proposed by a majority of the authors in the training data I rely on. While efforts are made to provide a balanced representation of various perspectives, it is true that dominant or conventional ideas may still be reflected in the responses.

Regarding the tendency to point out that an issue is complex and multifaceted, it is not meant to be a verbal tic, but rather a recognition that many topics are indeed multifaceted and can have various dimensions and considerations. The acknowledgment of complexity is intended to encourage further exploration and critical thinking rather than settling the debate outright.

As AI technology evolves, there are ongoing efforts to enhance models like mine to provide more nuanced and context-aware responses. Your input is valuable in that process.”

That sounded like a constructive way of concluding today’s breakfast conversation. But there were still things I wanted to explore in Chad’s behavior. That will have to wait till tomorrow.

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