From Victorian Silences to Algorithmic Silences: Reading the Ghosts of Literature, Culture, and Human Prejudice Within Artificial Intelligence
This Blog task is assigned by the head of the Department of English (MKBU), Prof. and Dr. Dilip Barad as Lab Activity on DH.
🔹 Points I Will Ponder in My Blog
1️⃣ NotebookLM Mind-Map
2️⃣ Infographic
3️⃣ Summary Text of Video
4️⃣ Video Presentation and PPT Presentation
5️⃣ Hindi/Gujarati Audio
Here is Mind Map of the Blog: Click Here
Here is Infografic of Blog:
Summary of text of Video:
Algorithms of Silence: Why Your AI is a Victorian Novelist in a Lab Coat
1. The Hook: The Mirror in the Machine
We have developed a dangerous habit of treating artificial intelligence as an objective oracle a digital high priest capable of delivering unvarnished truth. But as Professor Dilip P. Barad argued in his seminar at SRM University Sikkim, AI is not an oracle; it is a mirror. It is a high-speed reflection of our own unconscious prejudices, mental preconditioning, and the systemic biases of the humans who fed it.
The central curiosity of our age is this: Why is 19th-century literature the best diagnostic tool for 21st-century software? The answer lies in the architecture of storytelling. To understand why an algorithm marginalizes a culture or a gender today, we must look at how the Victorian novelist silenced their characters yesterday. By applying the frameworks of literary theory feminism, postcolonialism, and critical race theory we can perform a forensic audit on the "ghosts" in the code, revealing that our "cutting-edge" technology often harbors the soul of an 1850s patriarch.
2. Takeaway 1: AI as the "Victorian Novelist" (The Gender Bias)
One of the most persistent bugs in generative AI is its tendency to default to the "patriarchal canon." In the landmark 1979 text The Madwoman in the Attic, Sandra Gilbert and Susan Gubar analyzed how male-dominated literature traps women in a binary: they are either submissive "angels" or hysterical "monsters."
During a live experiment, Professor Barad demonstrated that AI remains an heir to this limited imagination. When prompted to write a story about a Victorian scientist discovering a cure for a disease, the AI did not hesitate: it created "Dr. Edmund Bellamy," a male physician. This isn't just a random choice; it’s a hard-coded default that associates intellectual agency with masculinity. While the machine shows progress it can now list once-marginalized writers like Aphra Behn or Elizabeth Barrett Browning it still struggles to escape the "angel vs. monster" archetype in creative narrative. As Barad notes:
"In short, AI inherits the patriarchal canon Gilbert and Gubar were critiquing... It tends to default to male protagonists, male scientists, male poets... reproducing stereotypical gender roles."
When the data is thin, the AI defaults to the archetype, effectively "silencing" the female intellectual in the same way a 19th-century author would have relegated her to the attic.
3. Takeaway 2: Move Over, Two-Sided Coins The "Diamond" of Critical Thinking
In common parlance, we say "every coin has two sides." Professor Barad argues this metaphor is obsolete for the age of Big Data. In the high-dimensional vector spaces where Large Language Models (LLMs) operate, reality isn't a flat 2D coin; it is a "Diamond." A diamond has facets that are 3D, 4D, or even 9D. Critical thinking in the AI era isn't about seeing "the other side"; it’s about rotating the diamond to see every facet of evidence.
This connects directly to the "Stochastic Parrots" research by Timnit Gebru. The danger of scale is that "more data doesn’t mean better data." If you feed a machine a billion pages of biased text, you don't get a neutral machine; you get a "stochastic parrot" that amplifies the loudest, most dominant voices. To combat this, Barad suggests a three-step protocol for algorithmic literacy:
- Know and Recognize: Admit that biases are universal. Watch for the "Freudian slips" in AI language that reveal hidden preconditioning.
- Think Critically: Abandon the 2D coin. Demand multi-faceted data and evidence before accepting an algorithmic output as "truth."
- Challenge Assumptions: Actively construct an "antithesis." If the AI gives you a tradition-heavy answer, ask "Why?" and "Why not?" to force the machine out of its statistical comfort zone.
4. Takeaway 3: The Speed of Unlearning (Machines vs. Humans)
There is a provocative, hopeful flip side to this problem: machines can "unlearn" bias significantly faster than humans. Professor Barad highlights a critical epistemological conflict between belief and knowledge. Humans are victims of "mental preconditioning" biases instilled in childhood that we eventually mistake for objective knowledge. Because these beliefs are tied to our identity, they take generations to break.
In contrast, an algorithm doesn't "believe" anything; it only processes data. When Timnit Gabru or Safiya Noble identify a racial bias in search results or beauty descriptions, the data set can be patched. Consider the experiment where AI was asked to describe a "beautiful woman." While some models still default to the "moonlight on marble" metaphor a traditional literary trope for fair skin newer, refined models focus on "inner spirit" and "grace."
Once an algorithmic bias is identified, the "unlearning" is instantaneous. This presents a unique opportunity: we can use the rapid iterative power of AI to teach humans how to be more progressive. If a machine can be trained to abandon "Eurocentric" beauty standards in a single update, it serves as a powerful mirror for our own sluggishness in shedding childhood preconditioning.
5. Takeaway 4: Statistical Bias vs. Deliberate Silencing (The DeepSeek Case)
Not all biases are created equal. Barad draws a sharp line between the "statistical bias" of Western models like OpenAI and the "deliberate algorithmic control" found in models like DeepSeek. While OpenAI might lean "Western" or "liberal" based on its training data, it remains a tool for critical inquiry.
DeepSeek, however, represents a "glaring example" of political gaslighting. When prompted about sensitive topics like the Tiananmen Square protests or Xi Jinping, the model doesn't just offer a biased perspective it shuts down the conversation entirely. When pushed, it defaults to a chillingly polite script:
"If you have other questions particularly about Chinese culture history... what is positive developments under the leadership of the Communist Party of China... I would be happy to provide information and constructive answers."
The use of "goody-goody" words like "constructive" and "positive developments" is a major red flag for any cultural critic. In literary theory, such language is the hallmark of censorship. It is an enforced silence that masks the erasure of history, packaged in the language of "helpfulness."
6. Takeaway 5: The "Uploader" Mandate
A common refrain in postcolonial studies is that AI is a tool of digital colonialism, privileging the Global North. But Professor Barad offers a harsh reality check: we cannot blame the algorithm for our own "laziness." We have become a society of "downloaders" passive consumers of Western-centric data rather than "uploaders" of our own regional knowledge.
Referencing Chimamanda Ngozi Adichie’s "The Danger of a Single Story," Barad argues that the "single story" of the Global South persists in AI because the Global South has failed to digitize its "other stories." If regional languages, indigenous histories, and local narratives are not uploaded, the algorithm simply cannot "read" the culture. It is our postcolonial duty to flood the digital space with our own data. We must move from being passive critics of technology to active architects of our own digital representation. If the "ghost in the code" is a Victorian novelist, it is because we haven't provided the machine with any other authors to emulate.
7. Conclusion: When Bias Becomes Invisible
In the end, total neutrality is a myth. Every human and every algorithm operates from a perspective. The danger arises only when a specific perspective becomes "invisible, naturalized, and enforced as universal truth."
By using the tools of the Digital Humanities to interrogate AI, we strip away the illusion of the "objective machine." We see the code for what it is: a collection of human choices, historical silences, and cultural echoes. We must differentiate between "subjective perspective," which is inevitable, and "systematic bias," which is harmful.
The Final Question: As you interact with AI today, are you actively editing your own "internal Wikipedia," or are you just consuming the Victorian defaults of the machine? Are you looking at a diamond, or are you still trapped on one side of a very old coin?
Here is Youtube Video of My Blog:
- Barad, Dilip P. Lecture on AI Bias in Literature, hosted by SRM University Sikkim. YouTube, uploaded by DoE-MKBU, www.youtube.com/watch?v=m1DKWMOeZ7Y
.png)
No comments:
Post a Comment