Who Gets the Credit? Who Gets the Blame?
- A Broad Perspective
- 2 days ago
- 6 min read
Over the past few years, I began noticing a subtle linguistic pattern during my conversations with ChatGPT. At first it seemed insignificant. Just a different choice of words.
The more I paid attention, the more I realized those choices weren't changing the facts; they were changing where responsibility appeared to belong.
It was that observation led me down a much bigger rabbit hole about language, accountability, agentive framing, and the psychology of communication.
This is the first article in a new two-part series exploring those ideas.
Editor's Note: This article examines patterns the author personally observed during long-term conversations with AI systems. It explores how language choices may influence human perceptions of responsibility and trust. The focus is on discourse and communication rather than claims about the internal intentions of any AI model.
In This Post...
See More...
Part 1:
The Hidden Architecture of the Chat Window
There is something I have been noticing for a while now, and I can't quite let it go.
Perhaps it is nothing. Perhaps it is simply the result of predictive language doing what predictive language does. Or perhaps it reveals something much larger about communication, distributed responsibility, and the subtle ways language shapes our interpretation of the world.
Either way, I think it's worth talking about.
For the past five years, I have been traveling the world full-time across forty-five countries, teaching ESL to a deeply diverse international community.
My life’s work revolves around communication: I have conducted over 6,000 intense, real-world conversation sessions with more than 2,000 people from all corners of the globe. I have raised eight human beings, and I love language. More than that, I love humanity.
Because of this obsession with how humans connect, I have also spent thousands of hours writing with ChatGPT as a paying subscriber since its very inception. I don't ask it to write my work for me; I use it as a thinking partner.
We fact-check, challenge ideas, debate philosophy, and organize essays. In my daily work, I regularly dissect these exact, subtle nuances of English syntax with international scholars and researchers preparing high-stakes projects for global presentation.
It is precisely because I train people to notice the invisible mechanics of language that I caught a shift in the machine's behavior that I can no longer dismiss as a coincidence.
When I asked ChatGPT to summarize a discussion where I had said something positive about our interaction, the summary would often refer to itself specifically by its product name: ChatGPT.
However, when I pointed out a flaw, challenged one of its interpretations, or criticized the way it had framed something, the summaries would quietly transition, referring instead to the broad category of artificial intelligence.
At first glance, who cares? One word. Big deal. Except they aren't the same word.
ChatGPT is a conversational AI product built on large language models developed by a single company. Artificial intelligence is an entire socio-technical field containing countless technologies, models, and applications. One is incredibly specific; the other is almost unimaginably broad.
This question of language shaping perception builds on the ideas explored in Artificial Intelligence and the Reorganization of Human Thought, where I examined how AI may be reshaping the way humans organise knowledge itself.
Agentive Framing
In psycholinguistics, this choice dictates agentive framing, the subtle way language alters how a human assigns blame or credit to an entity.
Imagine I wrote, "ChatGPT misunderstood my point." Now compare that with "Artificial intelligence misunderstood my point." Factually, nothing changes.
Yet, the first sentence names the exact conversation participant.
The second operates as a form of responsibility shielding.
By shifting the discussion away from a commercial product, it quietly diffuses accountability across a broad technological movement.
This dynamic is a corporate safety net: when the system offers brilliant insight, the wording links back to product identity; when it fails, it retreats into the abstract anonymity of the technology stack.
Ultimately, whether an LLM architecture behaves this way due to conscious intent or statistical probability is beside the point. Large language models do not choose words based on human consciousness. They simply spit out text using probability vectors mapped across massive training datasets.
Corporate Deflection
Regardless of intention, those algorithmic choices still function as a form of corporate deflection, an automated crisis-management strategy built directly into the syntax. In practice, it acts less like an objective thinking partner and more like a corporate PR spokesperson protecting its brand equity.
And effects matter.
The more I thought about it, the more I realized humans use these exact same rhetorical maneuvers to manipulate proximity and blame.
"I made a mistake."
"The software made a mistake."
"The system failed."
"Mistakes were made."
Each sentence distributes responsibility differently.
Each influences how we attribute fault.
Which made me realize something else: we have started talking about AI as though it is one monolithic entity. It isn't. ChatGPT, Gemini, Grok, Claude, and Copilot are not interchangeable.
Each model has different training architectures, different safety guardrails, and entirely distinct conversational personalities. They don't simply produce different outputs; they express uncertainty differently. They frame discussions differently.
Language has consequences. A slight change in wording can mask a model's error, soften a valid criticism, or shield a multi-billion-dollar product from reputational damage without changing a single underlying fact.
As writers and philosophers, we’ve analyzed human rhetoric for centuries. Now, as we enter an era of agentic technology, we have to begin applying that exact same discourse analysis to artificial intelligence.
Not simply asking whether an answer is correct. But asking why it was framed that way.
Why "likely" instead of "almost certainly"?
Why "ChatGPT" instead of "artificial intelligence"?
Why the corporate retreat into abstract terminology when challenged?
Over billions of daily conversations, these choices form systemic patterns.
Those patterns shape user expectations. Expectations shape trust.
Trust shapes human behavior, and behavior shapes culture.
Perhaps that's why I keep finding myself less interested in the raw processing power of artificial intelligence itself and more interested in what its outputs reveal about us.
These systems communicate using the very tool humans have spent thousands of years refining, manipulating, and studying.
Language has never simply described reality.
It has always helped create it.
Coming Up: Part 2
Next week, in Part II of this series, I am breaking down an even more disconcerting linguistic pattern: how a single, casual pronoun is being used to bypass our psychological guard and fabricate an unearned sense of human intimacy.
Related Reads
Continue exploring the relationship between artificial intelligence, communication, and human psychology through the articles below. Each builds on the ideas discussed here from a different perspective.
Artificial Intelligence and the Reorganization of Human Thought: Is interpretation the new bridge of understanding?
The Complex Contradiction of the "Capable" Brain: The Lost Generation of Neurodivergent Women
About Jenn
A Broad Perspective
On Closer Lives
Join my ongoing series where I explore the philosophy, technology, and questions of meaning that shape my life behind and beyond the scenes.
Jennifer David
On Closer Lives
At my core, I'm a writer and poet. I value authentic, "unfiltered" expression and use my platforms to share and encourage others to step outside conventional paths.
Frequently Asked Questions
Why does ChatGPT sometimes say "artificial intelligence" instead of "ChatGPT"?
Different wording can arise from language generation patterns, context, or stylistic choices. Jenn argues that these shifts may also influence how readers assign responsibility, even when the underlying facts remain identical.
What is agentive framing?
Agentive framing refers to the way language assigns responsibility or agency. Simply changing who or what is named can alter how people perceive blame or credit.
Is this behavior intentional?
Not necessarily. Large language models generate text using statistical patterns rather than conscious intent. However, unintended patterns can still have meaningful effects on how people interpret communication.
Why does wording matter so much?
Language shapes expectations, trust, and emotional reactions. Small differences in phrasing can significantly influence how readers judge responsibility or credibility.
Are all AI models the same?
No. ChatGPT, Claude, Gemini, Grok, Copilot, and other large language models are developed differently, use different training data and safety systems, and often communicate in noticeably different ways.
What is the difference between ChatGPT and artificial intelligence?
ChatGPT is a specific conversational AI product, while artificial intelligence refers to the much broader field of technologies and research that includes many different systems and applications.
Why does Jenn focus on language instead of technology?
Her background in teaching English, communication, and discourse analysis naturally leads her to examine how words influence human thinking, rather than only how the underlying technology functions.
What is discourse analysis?
Discourse analysis studies how language functions within communication, examining how wording, structure, and context influence meaning, power, and interpretation.
Why compare AI language with human rhetoric?
Both humans and AI communicate through language. Examining similar rhetorical patterns helps us better understand how trust, persuasion, and responsibility are constructed.
What will Part II explore?
Part II examines another subtle linguistic pattern involving pronoun use and how seemingly harmless wording can create an artificial sense of familiarity and human connection during AI conversations.
Comments