Lab meeting presentation of Proctor 2026 Authoring

# Authoring Dialogic Identities in Collaboration with Artificial Intelligence

Computational Literacies Lab meeting

February 18, 2026

]

???
International Journal of Computer-Supported Collaborative Learning.
Supported by NSF Grant No. 2219433.

---

# Thanks, Claude

```
Create a document, `2026_ijcscl_slides.md`, which creates a slide presentation 
based on the article in `2026_ijcscl.md`, for an academic audience, to present 
for about 30 minutes. These slides will be implemented using remark.js.

The file should only contain a `<textarea id="source">` with remark.js slide 
content—it will be injected into a template (`site/themes/lab/templates/slides.html`),
which makes available CSS from `site/themes/lab/static`. This is the only CSS styling 
available. Do not use partial reveals on slides.
```

Evaluation: Created a reasonable first draft. Required heavy revision.

---

## A foundational question for CSCL

As generative AI becomes embedded in everyday writing practices, To what extent can interaction with large language models constitute ***collaboration***?
What are the implications for ***authorship***, ***identity***, and ***agency***?

We argue that a dialogic understanding of identity is essential for conceptualizing collaborative writing with AI.

> CSCL conceptualizes computers as tools that *mediate and support* collaboration—but not as primary *participants* in collaboration.

---

# Research questions

1. **How is authorship distributed** in collaborative writing with AI?

2. **How does the dialogic nature of authors' identities** shape the writing process?

---

# Background

---

# Writing as social action

- All writing is **social**: even writing alone involves an imagined audience .cite[1]
- Writing constitutes **literacy practices**—social practices mediated by language use .cite[2]
- CSCL has studied writing as a collaborative task, but emphasis is almost always on **epistemology** rather than **rhetoric or communication**
- We argue for a **rhetorical and dialogic approach**: *Who is speaking here? Who is being addressed? Where does this language come from?*

---

# Three authorial roles

| Role.cite[1] | Definition |
|------|-----------|
| **Animator** | The person who inscribes the words on the page |
| **Author** | The person who selects the sentiments and ideas |
| **Principal** | The person who is committed to what the words say |

These roles illuminate why AI is accepted in some contexts and rejected in others:

- **Email drafting**: AI as animator (accepted—like a greeting card)
- **Heartfelt public statements**: fraudulent authorship (betrayal)
- **Romantic chatbots**: ersatz principal (deeply controversial)

---

# Dialogic identity

- **Dialogism**: meaning is constituted through ongoing interaction among multiple voices .cite[1]
- The words we use have **histories**; we assemble them in relation to prior uses and in anticipation of future audiences
- Identity is a model of selfhood one **authors and occupies** within a literacy practice .cite[2]

> "An identity depends on the individual's understanding of how that identity will be recognized in that relationship, time, or context."
> —Moje & Luke, 2009

- AI **models language, culture, and discourse**—when we interact with AI, our identities and agency become intertwined

---

# Theoretical contribution

A **dialogic understanding of identity** is essential for conceptualizing collaborative writing with AI.

- CSCL's existing engagement with dialogism focuses on **distributed knowledge production**
- Less attention to how dialogism shapes **who we can be** (identity) and **what we can do** (agency)
- AI is unlike earlier educational technologies: it capably participates in the same **cultural discourses** we use to anchor our identities

---

# New Interfaces for Collaborative Writing with AI

---

# Unfold Studio

- A **web-based platform** for reading and writing interactive stories

- Stories written in **Ink** (Inkle, 2019): a programming language giving primacy to text

- Reader cycle: text → choices → text → choices → end

## Declarative narrative

Just as *declarative programming* expresses intent and lets the computer figure out details, **declarative narrative** delegates components of text to generative AI while retaining overall control.

---

# Three AI affordances

| Affordance | Delegates | Authorial role delegated |
|-----------|-----------|------------------------|
| **`generate`** | Sensory detail | **Animator** |
| **`continue`** | Plot progression | **Author** |
| **`agent`** | Characterization | **Principal** |

These affordances **make visible** dynamics that already exist in prose in more subtle forms—a benefit for pedagogical applications.

---

# `generate`: Delegating detail

- Generates text in response to a provided prompt
- Similar to ChatGPT, but control flow returns to the story

```
{generate("One sentence describing an alien ship's
long halls with three doors.")}
```

→ *"The hallway's endless metal panels gleamed under cold violet luminescence, and three silent doors—one etched with shifting glyphs, one glowing with liquid emerald light, and one veiled in crackling blue plasma—stand like sentinels..."*

---

# `continue`: Delegating plot

- Generates a series of **open-ended user inputs** followed by AI responses
- Requires a **target knot**: the point where the scripted story resumes
- AI guides the interaction toward the target; rejects implausible inputs

```
-> continue(-> ending5)
```

The author retains the **principal** role by explicitly bounding where the open-ended interaction starts and ends.

---

# `agent`: Delegating character

- Requires **target** (story resumption point) and **character** (a sub-story modeling the character)
- AI generates responses **in the voice** of the character
- Makes it possible to write stories enacting a **dynamic struggle for control** between authorial voice and characters

---

# Methods

---

# Study context

- **Ten-week after-school writing club** at a high school (northeastern US)
- Led by the first author
- Structured as a **writer's workshop** .cite[1]

**Data sources:**
- Sequence of **story versions** (every save produces a version)
- Records of **authors' playthroughs**
- Facilitator's **fieldnotes** and ongoing discussions
- Pre- and post-**surveys**

Story edits classified as: **Debugging** · **AI** · **Code** · **Text**

---

# Analytical approach

- **Literary analysis** of how two focal authors developed stories
- Focus on:
  - How authors use familiar literary elements *and* new AI affordances for rhetorical effect
  - How they inhabit the three authorial roles (animator, author, principal)
- **Contrasting cases** of dialogic authorial identity

---

# Case Studies

---

# Case 1: Dragonbaby

- 9th-grade girl; collaborates with her 12th-grade sister (pair-programming style)
- Avid readers of fantasy; enjoy **fan-fiction**
- Final story: *"Our actions have consequences"* — fan-fiction set in *Percy Jackson*
- Also worked on *"Worlds collide"* (Percy Jackson × Harry Potter)
- **Neither had used generative AI before**

---

# Dragonbaby: Story development

- **Early phase**: intense debugging + code edits (learning the programming language)
- **Middle phase**: intense text and code edits with decreasing debugging
- **Later phase**: AI affordances added after most text/code in place
- AI edits triggered an **intense period of playthroughs**
- By the end: no longer editing—only **repeatedly playing** the story
- Continued playing **weeks after the workshop ended**

---

# Dragonbaby: Prompt iteration

Development of a single `generate` prompt across four versions:

| # | Prompt |
|---|--------|
| 1 | *a discription of the seven introducing themselves to Danielle and Draco* |
| 2 | *...the seven **plus Grover** introducing themselves...* |
| 3 | *...the seven **from Heroes of Olympus** plus Grover...* |
| 4 | *...to **Danielle Malfoy** and Draco* |

Each version interspersed with playthroughs examining generated text before tweaking.

---

# Dragonbaby: Authorial roles

- **Animator** → frequently delegated to AI via `generate` and `continue`
- **Author** → increasingly *shared* with AI
  - Delighted to discover the LLM knew Percy Jackson lore
  - AI as co-author, dynamically deciding what happens next
- **Principal** → clearly retained ("her story"), but AI gained a subtle principal role

> Through intensive replaying and geeking out, Dragonbaby authored a **positive dialogic identity** embedded in AI-generated discourse—a fan-fiction literacy practice shared *with* the AI.

---

# Case 2: Quaezae

- 10th-grade girl; quiet, preferred working alone
- Very interested in **computational aspects** of interactive stories
- Expressed **moral opposition** to generative AI
  - *"GAI was ruining her friends' minds"*
  - Only used AI on friends' phones so it wouldn't be connected to her
- Agreed to participate after discussing how the technology worked
- Invited to **experiment with AI to become better-informed in her opposition**

---

# Quaezae: "Tellers"

A **dystopian story** in which implanted AI chips are required to participate in society.

> "AI has completely been enthralled into our lives... In this world, its been complete accepted, or more accureately ... **integrated**."

The distinction between "integrated" and "accepted" perfectly articulates the **dialogic nature of AI**.

---

# Quaezae: Two endings

**Accept the chip** ("The Good Ending" — sarcastically glossed):

> "Welcome back, Eric—how does it feel to be more than human? [...] you notice how…simple it is to have the new implant in your arm"

**"Reject modernity!"** ("Neutral Ending"):

> "You go on with your life like normal... but less and less places accept your physical card... You begin going down a path of stealing and crime"

An example of **interactional positioning** .cite[1]: the choice between giving yourself up to AI or social exclusion.

---

# Quaezae: Authorial roles

- **Animator** → delegated to AI via `generate`
- **Author** → *not shared* with AI (wary, reluctant engagement)
- **Principal** → retained, but AI present in an **adversarial** mode
- Asks for reader's **name**, then injects it into AI prompts → unsettling experience of being put into dialogue with AI

> Quaezae's authorial identity is also **dialogically enmeshed with AI**—but *against her will*. The story feels claustrophobic: the authorial voice rejecting something she recognizes as part of herself.

---

# Discussion

---

# Two contrasting dialogic identities

A **dialogic account of identity** is essential for conceptualizing and understanding collaboration with generative AI.

| | Dragonbaby | Quaezae |
|---|-----------|---------|
| Stance | Inspired, validated | Creeped out, "enthralled" |
| Mode | Enthusiastic co-authorship | Resistance and critical distance |
| AI relationship | Geeking out together | Fearing unwanted integration |

Both recognize the **dialogic nature** of writing with AI. Both author identities that are **distributed in literacy practices shaped by AI**.

---

# AI as active collaborator

- AI-based media is better understood as an **active collaborator in meaning-making** than as a tool passively mediating collaboration between people
- Instead of gaining control over AI as we internalize it, we **co-author our identities** and **co-produce meanings** with AI
- If collaboration may **transform us**, then collaboration with AI has the potential for a **sinister vulnerability without reciprocity**

---

# Dialogic criticality

- AI motivations and biases are **opaque**, shaped by commercial and geopolitical interests
- Need for **dialogic criticality**: not just interrogating AI as an external entity, but **introspecting on our own authorial identities** that are becoming intertwined with AI
- Interactive storytelling has particular potency as a tool for **dialogizing AI** while remaining in control of the interpretive context

> The novel destabilizes other literary forms so they "become dialogized, permeated with laughter, irony, humor, elements of self-parody..."
> —Bakhtin, 1981

---

# Conclusion

> Just as social media dialogized, colonized, and transformed our social relationships, AI is becoming ubiquitous in our literacy media and stands to **dialogize our authorial identities**.

- We are in the same situation as the reader in "Tellers": opting in means being transformed in ways we cannot choose; opting out means exclusion
- A dialogic understanding of identity can help bring into focus the **mechanisms and consequences** of our relationships with AI
- Technologies and pedagogies like those explored here may help create **critical possibilities for living in the age of AI**

---

# Thank you

Chris Proctor · Varun Bhatt · Ryan Rish

chrisp@buffalo.edu

Unfold Studio: unfold.studio

NSF Grant No. 2219433