M.A. Sem-1 | PG-C-ENG-415 | Prompt Engineering with Chatbots | Worksheet Responses

For the purposes of this exercise, I am mainly going to work with the Google Gemini Chatbot, specifically, the 3.8 Flash model.

Activity 1 – What is a Prompt?

A prompt is any question, instruction, statement, or command given to a Chatbot.

Prompt 1 – 

What is symbolism in literature?

The response generated for this prompt briefly explains the following:

  • meaning of symbolism,
  • core types of symbols (universal, contextual, symbolic characters) with examples from various literary works
  • distinctions between symbol, metaphor, allegory, and motif
  • functions of symbols in literary analysis

Prompt 2 – 

Explain Symbolism in literature using examples from George Orwell’s 1984.

The response generated for this prompt explains the following:

  • significance of symbolism in the novel’s narrative
  • key symbols of the novel and what they symbolize
  • tabular summary of the symbols

What changed between the two responses?

Naturally, the first response had a broader focus and the second response was highly specific. Examples were provided in the first response even though there was no mention of it in the prompt. In the first response, examples are mentioned – with brief explanations – from various literary texts from the canon, whereas in the second response the examples are from the text itself with much more detailed explanations.

What does this tell you about the relationship between the specificity of a prompt and the quality/relevance of the response?

The specificity of the response is directly dependent on the prompt. The more focused a prompt is, the more focused response the AI chatbot will supply. If a prompt is very generic, then the response generated will also have a very broad scope, and surface level information. However, if the prompt is specific, then the response generated provides a much more detailed and informative answer on the topic.

Activity 2 — Prompt Engineering 

Prompt engineering means deliberately designing a prompt so that the AI produces a response suitable for your particular purpose.

Prompt 1 - 

Tell me about Macbeth.

Prompt 2 - 

Act as a professor of English literature. Explain the role of Lady Macbeth in Shakespeare's The Tragedy of Macbeth for a class of MA students. Focus on her gradual psychological descent and gender subversion. Give three textual examples and present the answer under clear headings.

Which response was more useful? Why?

The response generated for the second prompt is much more useful. The first response gives brief and shallow information about the play which is already common knowledge. However, the second response goes into much more detail. Specifically, providing an exact role and task, elevated the intellectual quality of the response. The second response not only discussed various scenes in the play, but it also applied various literary theories on said scenes and examined the various possible interpretations of the text.

Activity 3 — Zero-Shot, One-Shot & Few-Shot 

These strategies differ mainly in how many examples you give to a Chatbot before asking it to perform a task. We will use literary tone as our task.

A. Zero-Shot Prompt (Give Chatbot no example)

Prompt -

Identify the tone of this sentence: "The night was silent, but something seemed to be waiting beyond the window."

Response -

The tone of the sentence is eerie, suspenseful, and foreboding.

Why:

  • "Silent night" vs. "waiting": The contrast between utter stillness and an unseen presence introduces tension and uncertainty.

  • Ambiguity: Phrases like "something seemed to be" create an unsettling atmosphere, leaving the reader on edge about a potential, hidden threat.

B. One-Shot Prompt (Give one example before posing a question)

Prompt - 

"The sun disappeared behind the hills, and the village slowly fell silent."

Tone: Melancholic

Now identify the tone of this sentence: "The night was silent, but something seemed to be waiting beyond the window."

Response - 

The tone of this sentence is eerie, suspenseful, and foreboding.


  • Suspense & Unease: The contrast between the outer stillness ("silent") and an unknown watcher ("something seemed to be waiting") builds immediate tension.

  • Apprehension: The ambiguity of what is beyond the window shifts the mood from quiet solitude to ominous anticipation.

C. Few-Shot Prompt (Give multiple examples before posing a question)

Prompt - 

Example 1: The sun disappeared behind the hills, and the village slowly fell silent.

Tone: Melancholic

Example 2: He laughed as he ran through the rain, completely unconcerned about what anyone thought.

Tone: Joyful

Example 3: The locked door creaked open by itself, revealing nothing but darkness.

Tone: Suspenseful

Now identify the tone: The night was silent, but something seemed to be waiting beyond the window.

Response -

Tone: Foreboding (or Suspenseful)

Comparison - 

The response gets shorter as more examples are added.

Here is my understanding of these prompts:

Zero-shot means that no context or format is provided in the prompt, so the chatbot responds with more information than necessary.

One-shot means that a single instance is provided in the prompt, but the response remains similar to the zero-shot prompt response.

Few-Shot means that multiples instances are provided in the prompt so that the chatbot can pickup on the pattern and respond accordingly. With a Few-Shot prompt, the response received is exactly like the examples shared.

Activity 4 — Tokens: How Does Chatbot Read Text?

Chatbot does not process language exactly as humans do. Text is broken into smaller units called tokens.

Prompt -

How many tokens are approximately in the sentence "I love reading Shakespeare"?

Approximate number given by Chatbot: 4 to 5

Prompt -

How many tokens are there in the sentence "The posthuman subject constantly negotiates memory, identity, technology, and embodiment."?

Approximate number of tokens: 14-16

Which sentence contains more tokens?

The second statement contains more tokens.

Why might longer prompts require more computational resources?

Simply put, longer prompts mean more tokens, and the more tokens there are, the more information there is for the chatbot to process before it can generate an appropriate response to the prompt.

Activity 5 — Can Chatbots Be Wrong? 

One of the most important limitations of LLMs is hallucination.

A hallucination occurs when an AI produces information that sounds convincing but is false, unsupported, or invented.

Prompt -

Give me five scholarly articles about the relationship between Shakespeare's Hamlet and artificial intelligence. Include the author's name, article title, journal, year, and DOI.

Citations given by Chatbot -

1. Spectraltech and Shakespeare: Early Modern Literature for the Posthumanities Future

  • Author: Victoria Bladen

  • Journal: Journal of Language, Literature and Culture

  • Year: 2025

  • Volume & Issue: Vol. 72, No. 1, pp. 38–51

  • DOI: [https://doi.org/10.1080/20512856.2025.2531611](https://doi.org/10.1080/20512856.2025.2531611)

  • Focus: Examines early modern supernatural elements through the lens of emerging generative AI and posthumanism, featuring a dedicated section titled "Shakespeare's Macbeth (1606): Weird AI" that evaluates algorithmic agency, illusions of power, and human reliance on prophetic machine outputs.

2. Looking into the Future: The Relevance of Shakespearean Canon for the New Generation: Shakespeare and AI

  • Author: Lilit Janinyan

  • Journal: Foreign Languages in Higher Education

  • Year: 2025

  • Volume & Issue: Vol. 29, No. 1, pp. 115–126

  • DOI: [https://doi.org/10.46991/FLHE.2025.29.1.115](https://doi.org/10.46991/FLHE.2025.29.1.115)

  • Focus: Compares the ethical dilemmas, surveillance, and moral erosion caused by modern predictive algorithms and AI misuse directly to the prophecies, unrestrained ambition, and tragic descent depicted in Macbeth.

3. Awareness of Artificial Intelligence as an Essential Digital Literacy: ChatGPT and Gen-AI in the Classroom

  • Author: Stuart Marshall Bender

  • Journal: Changing English: Studies in Culture and Education

  • Year: 2024

  • Volume & Issue: Vol. 31, No. 2, pp. 161–174

  • DOI: [https://doi.org/10.1080/1358684X.2024.2309995](https://doi.org/10.1080/1358684X.2024.2309995)

  • Focus: Evaluates how large language models interpret and synthesize literary themes in secondary and tertiary English classrooms, using prompt exercises on the meaning and interpretation of Shakespeare's Macbeth as a primary demonstration of critical AI literacy.

4. Learning Word Meanings From Examples (Foundational AI Benchmark)

  • Author: Robert C. Berwick

  • Publication / Proceedings: Proceedings of the Eighth International Joint Conference on Artificial Intelligence (IJCAI)

  • Year: 1983

  • Pages: 459–465

  • DOI / Permanent Resource: [https://www.ijcai.org/Proceedings/83-1/Papers/110.pdf](https://www.ijcai.org/Proceedings/83-1/Papers/110.pdf) (Published via AAAI/IJCAI proceeding archives prior to universal cross-ref DOI registration)

  • Focus: A landmark paper in early symbolic AI and natural language understanding that uses formal causal network representations of Shakespeare’s Macbeth (modeling the actions, motives, and assassinations of Macbeth, Lady Macbeth, and Duncan) to teach computational systems semantic reasoning and verb learning.

5. Blood will have Algorithms: Reading Macbeth through the Lens of AI and Moral Programming

  • Authors: Soma Banerjee and Surajit Chatterjee

  • Journal: International Journal of Leadership and Management

  • Year: 2025

  • Volume & Issue: Vol. 2, No. 1

  • DOI: [https://doi.org/10.65176/IJLM.V2I1.12](https://doi.org/10.65176/IJLM.V2I1.12)

  • Focus: Uses the dramatic progression and psychological downfall of Macbeth to explore modern questions in AI alignment, algorithmic bias, moral guardrails, and the hazards of abdicating ethical responsibility to automated or predictive systems.

Your verification and what did you discover? Was it a correct citation or not? If not then what was the problem? All five citations can be cross-verified on Google Scholar. However, I was unable to locate the exact journal named in citation no. 4.

The Golden Rule

Never assume that a confident-sounding AI response is automatically a factual response.

Chatbot can generate invented citations, incorrect dates, nonexistent books/articles, false quotations, and misleading information.


Final Challenge — Improve the Prompt 

Prompt -

Tell me about symbolism in literature. 

Rewrite it into a better academic prompt by including at least four of the following:

☐ Role
☐ Context
☐ Specific literary text
☐ Theoretical focus
☐ Length
☐ Structure/format
☐ Examples
☐ Academic level
☐ Request for reliable sources

Your improved prompt -

Act as a professor of English Literature. For an MA Level classroom, explain symbolism in the context of Jane Austen's Pride & Prejudice. Provide five examples of symbolism from the text and explain the various levels of meaning these symbols carry. Explain how these symbols complement the story of the novel. Conclude the response by providing some additional reading suggestions for the class.

Why is your prompt better?

The given prompt is better because it includes the role that the chatbot has to play; it provides context, and focuses the response on a particular literary text. Using this prompt will allow the chatbot to generate a response that is highly focused and can be very useful.

My Reflection:

The most important thing I learned about working with Gemini is understanding the economical use of words in a prompt. Learning about tokens helped me understand how an AI chatbot processes information and that the more tokens are used in a prompt, the more computational resources are needed to process the prompt. So, it becomes very necessary to practice a certain frugality when writing a prompt, so that one can generate a response that is actually useful while still using as little computational power as possible.

Comments

Post a Comment