The Literary Web: Mapping Meaning in the Digital Age

The Digital Quill

This Blog is a part of Lab Activity assigned by Dr. and Prof. Dilip Barad sir regarding Digital Humanities on how machine can help us write poems and an activity assigned regarding Dickens' Project and I will share my understanding of the same in this Blog.

What if Machines write poems?

In his talk, Oscar Schwartz presents the audience with several poems and asks them to determine whether each poem was written by a human or generated by an algorithm. This interactive exercise explores the increasingly blurred boundary between human creativity and artificial intelligence.

Successes and Failures

The results reveal that identifying the author is not always straightforward.

In some cases, the audience correctly recognizes poems written by human poets such as William Blake.

In other instances, algorithm-generated poems successfully deceive readers. Programs such as Racter (developed in the 1970s) and RKCP (designed to imitate the style of Emily Dickinson) convince a significant portion of the audience that their poems were written by humans.

These examples demonstrate that computer-generated poetry can sometimes closely resemble human literary expression.

The "Reverse Turing Test"

One of the most surprising examples involves a poem by Gertrude Stein. Rather than being recognized as human-written, the majority of the audience believed it had been generated by a computer.

Schwartz describes this as a "reverse Turing test," in which a human writer is mistaken for an artificial intelligence because of their unconventional writing style. This challenges our assumptions about what human creativity is supposed to look like.

Defining the Turing Test

The talk revisits the Turing Test, proposed by Alan Turing in 1950.

The Turing Test was designed to determine whether a computer could exhibit intelligent behaviour by engaging in a text-based conversation that is indistinguishable from one with a human.

According to the original idea, if a computer can fool a human judge approximately 30% of the time, it is considered to have passed the test. Schwartz points out that some poems in his "Bot or Not" database fooled readers as much as 65% of the time, suggesting that computers can sometimes imitate human creativity even more successfully than expected.

Philosophical Insights

Rather than focusing solely on whether computers can write poetry, Schwartz uses the experiment to raise deeper philosophical questions about humanity and artificial intelligence.

1. The Category of "Human" Is Unstable

Schwartz argues that being "human" is not a fixed or objective category. Instead, our understanding of what counts as human changes over time and is shaped by cultural expectations, beliefs, and interpretations.

2. Poetry as a Benchmark of Humanity

Poetry has traditionally been regarded as one of the highest forms of human creativity and emotional expression. Therefore, asking whether a computer can write poetry is ultimately another way of asking:

What does it truly mean to be human?

3. The Computer as a Mirror

Schwartz describes artificial intelligence as a mirror rather than an independent creator.

A computer reflects whatever idea of humanity we teach it. For example, if it is trained on the works of Emily Dickinson, it reproduces patterns, language, and stylistic features that resemble Dickinson's poetry. In this sense, AI reflects human culture back to us rather than inventing something entirely new.

Conclusion

Schwartz concludes that the most important question is not simply whether artificial intelligence can become human-like. Instead, the more profound question is:

What vision of humanity do we want artificial intelligence to reflect back to us?

As AI becomes increasingly capable of producing creative works, it challenges us to rethink our definitions of creativity, authorship, intelligence, and, ultimately, what it means to be human.

My Reflection: Identifying AI-Generated Poetry

After reviewing the collection of poems, I began to notice recurring patterns that often distinguished AI-generated poetry from human-written work. Although many of the AI-generated poems initially appeared convincing, a closer and more attentive reading revealed subtle differences in language, structure, and presentation.

Overall, I correctly identified five out of the six poems. I misclassified only one poem, assuming it had been generated by AI when it was, in fact, written by a human. This experience reinforced my understanding that while artificial intelligence can imitate poetic expression remarkably well, careful reading still exposes certain inconsistencies.

One of the most noticeable differences lay in word choice and sentence structure. AI-generated poems often followed recognizable linguistic patterns that felt highly organized and predictable, whereas human poetry tended to display greater spontaneity, ambiguity, and stylistic individuality. These differences became more apparent through close reading rather than a quick first impression.

I also observed distinctions in punctuation, formatting, and stanza arrangement. Human-written poems frequently employed punctuation and line breaks in ways that reflected a poet's personal rhythm and intention, often appearing in one or two cohesive stanzas. By contrast, many AI-generated poems were divided into multiple evenly structured stanzas with a more uniform visual layout. While these features are not universal, they served as useful indicators during the exercise.

This challenge demonstrated that understanding even the basic characteristics of AI-generated writing can help readers critically evaluate a poem's authorship. Rather than relying solely on intuition, paying attention to linguistic patterns, formatting choices, and poetic style allows us to make more informed judgments. At the same time, the one poem I misidentified served as a reminder that the boundary between human and machine creativity is becoming increasingly blurred, making close reading more important than ever.

Here are the images of the quiz-

CLiC Activity-

10. Setting and atmosphere in novels

For the creation of fictional worlds, the setting and atmosphere play an important

role. While each novel creates its own particular world, it is still possible to identify

similarities across novels and we can interpret accounts of settings against the social

and historical context of the time. Charles Dickens is often referred to as an author

who was concerned with living and working conditions in the city, Jane Austen, in

contrast often shows us social life away from the city. A starting point to compare

the type of fictional worlds that these two authors write about is a ‘key comparison’.

Link for the Drive of activity-

Austen Vs Dickens

Personal Learning Outcome

• This activity helped me understand how Digital Humanities and corpus-based tools can be used to approach literary texts in a more systematic and evidence-based way. Before working with the CLiC tool, I generally approached novels through close reading, where I focused on themes, characters, language, setting, and the overall meaning of the text. However, this activity introduced me to a different way of reading literature—by using computational analysis to identify patterns in language and then connecting those patterns with literary interpretation.
• While comparing the novels of Jane Austen and Charles Dickens, I learned how the concept of keywords can reveal significant differences between two authors. The keyword analysis showed that Austen's novels contained relatively more words associated with feelings, happiness, manners, behaviour, civility, marriage, invitations, dancing, and social relationships. These words provided an indication of how strongly Austen's fictional world is connected with interpersonal relationships, social conduct, courtship, and domestic life. On the other hand, Dickens's keywords included words such as streets, city, prison, money, door, fire, light, face, hands, and eyes, which pointed towards a more concrete and physically detailed fictional world, particularly one concerned with urban spaces, material conditions, characterisation, and social realities.
• An important learning outcome for me was realising that keywords themselves do not provide a complete interpretation. They only give us clues or patterns that need to be investigated further. For instance, finding words such as face, hands, or eyes frequently in Dickens does not automatically explain their significance. We need to examine their contexts through concordances and close reading to understand whether these words contribute to characterisation, atmosphere, symbolism, or the representation of physical reality. Similarly, the frequent appearance of words such as marriage or feelings in Austen invites us to explore how these concepts function within her narratives rather than simply assuming what they mean.
• Through this activity, I therefore learned to combine quantitative evidence with qualitative literary interpretation. The computer can identify recurring linguistic patterns, but the interpretation of those patterns still requires a human reader who can consider context, themes, historical circumstances, genre, and authorial style. This helped me understand that Digital Humanities does not replace traditional literary reading; rather, it can support and enrich close reading by directing our attention towards patterns that might otherwise remain unnoticed.
• Overall, this activity strengthened my understanding of how technology can be used as a literary research tool. It also encouraged me to look at familiar authors such as Austen and Dickens from a new perspective, where their distinctive fictional worlds can be explored not only through individual passages but also through broader patterns in the language of their novels.

Group Learning Outcome

• This activity was also a valuable experience in collaborative learning, as I worked together with Hiralba, Jaypal, and Divya to explore the differences between Jane Austen's and Charles Dickens's fictional worlds through keyword analysis. Rather than approaching the activity individually, we divided our attention towards understanding the CLiC tool, identifying significant keywords, comparing the results, and discussing what those words could reveal about the two authors.
• Working as a group helped us understand that literary interpretation can become more meaningful when different readers bring their own observations and perspectives to the same piece of data. As we examined the keyword lists, we identified several interesting patterns. In Austen's novels, words related to feelings, happiness, manners, behaviour, civility, kindness, marriage, dancing, invitations, and social relationships appeared prominently. In contrast, Dickens's novels showed keywords connected with body parts such as face, hands, eyes, and head, as well as concrete elements of setting such as doors, fire, light, glass, walls, water, streets, city, money, and prison.
• Discussing these findings together helped us move beyond simply collecting words. We began to ask why these particular words were significant and what they might tell us about the fictional worlds created by the two authors. We connected Austen's keywords with her representation of social relationships, interpersonal behaviour, courtship, marriage, and domestic life, while Dickens's keywords led us towards questions of urban life, physical surroundings, social conditions, characterisation, and the material environment.
• One of the most important things we learned as a group was that data requires interpretation. A keyword list by itself cannot provide a complete literary argument. We had to discuss the possible meanings behind the patterns and recognise that the same word can have different meanings depending on its context. This encouraged us to think critically rather than accepting the computer-generated results as final conclusions.
• The activity also improved our communication, cooperation, and analytical skills. Each member contributed observations and helped connect the computational findings with literary concepts such as setting, atmosphere, characterisation, genre, and authorial style. Through discussion, we were able to compare our interpretations, question each other's assumptions, and arrive at a more balanced understanding of the results.
• Most importantly, the activity showed us that Digital Humanities can be both an individual and collaborative practice. The use of a corpus tool gave us quantitative evidence, while our group discussion allowed us to transform that evidence into literary interpretation. Working with Hiralba, Jaipal, and Divya therefore helped me understand the value of combining technology, teamwork, critical thinking, and close reading in the study of literature.
• Overall, our collaborative work demonstrated that literary analysis does not have to remain limited to individual reading. When computational tools and different perspectives are brought together, they can open up new ways of understanding familiar authors and texts.

References-

Barad, Dilip. “What if Machines Write Poems.” Dilip Barad | Teacher Blog, 21 Mar. 2017, blog.dilipbarad.com/2017/03/what-if-machines-write-poems.html. Accessed 8 Aug. 2026.

Mahlberg, Michaela, Peter Stockwell, and Viola Wiegand. CLiC – Corpus Linguistics in Context: An Activity Book. Version 1, University of Birmingham, Nov. 2017, CLiC Activity Book. Accessed 8 Aug. 2026.

Mahlberg, Michaela, Peter Stockwell, Viola Wiegand, and James Lentin. CLiC 2.1: Corpus Linguistics in Context. 2020, CLiC. Accessed 8 Aug. 2026.