# bok-ai-lab-20250328

## **AI Lab Report – Week 1: Moving Beyond the \"AI Tutorbot\"**

In our opening session, we challenged conventional views of artificial intelligence in education, shifting the narrative from AI tutor models towards developing agentic and recursive workflows designed to augment existing classroom mechanics.

To set the stage, we began by reviewing recent news developments in AI, supported by a generated glossary of key terms to ensure participants from diverse academic backgrounds and varying levels of technical expertise could effectively engage with the content. Featured headlines included: "OpenAI Announces New Audio APIs," "Google Announces Gemini 2.5," and "Vibe Coding Goes Viral."

### Key Experiments and Discussions:

#### 1. Exploring AI \"Knowledge\" through Mathematical Reasoning

Participants engaged with ChatGPT in two distinct scenarios to critically assess its problem-solving capabilities:

- **Pure LLM Reasoning:** Users prompted ChatGPT to solve basic mathematical problems relying solely on its language model capabilities. This highlighted intrinsic heuristic limitations and "illusions of competence," making transparent both strengths and shortcomings in the way LLMs handle conceptual reasoning tasks.

- **Augmented with Coding Tools:** By enabling ChatGPT to leverage coding environments, participants could observe improvements in computational accuracy and logical reasoning, clearly delineating the boundary between linguistic inference and true computational capability.

This comparative exercise underscored critical insights into where AI can be reliably applied, as well as cautionary lessons about its limitations.

#### 2. Slack Glossary Bots: Dialectical Approaches to Definitions

We introduced a trio of innovative Slackbots, designed around the concept of generating recursive and dialectical definitions for user-submitted terms. Each bot provided distinct perspectives:

- **Neutral Bot:** Offered straightforward, unbiased definitions.
- **Friend Bot:** Generated definitions with positive connotations, framing concepts in an optimistic and supportive manner.
- **Foe Bot:** Presented definitions from a critical or challenging viewpoint, deliberately surfacing oppositional perspectives.

### Insights and Methodological Reflections:

- **Design Pattern – Expertise-Agnostic Applications:** Participants explored applying AI in educational moments that don't rely heavily on expert intervention—such as peer feedback sessions and paper workshopping. Identifying these "expertise-light" segments in pedagogical workflows allows for meaningful AI augmentation without compromising the quality or depth of human-centered interactions.

- Rosie vs. Roomba: Leveraging well-known references to classic AI-driven household robots (Rosie from "The Jetsons" and Roomba), we conducted interactive thought experiments designed to help participants differentiate between AI as a broad conceptual intelligence versus specific-purpose automation. This exercise helped clarify and crystallize understandings of AI's functional range.

- **Platform Strategy:** Our tools and experiments are primarily built for three platforms: Slack (interactive bot-driven workflows), Next.js (dynamic web-based interactions), and Google Colab (accessible coding environments). Participants expressed particular interest in detailed tutorials to facilitate practical mastery, notably Slackbot development and Google Colabs.