AI case study

 AI case study

Case Study 1:

Reimagining ELIZA for Modern-Day Mental Health Support in Underserved Communities

In the modern world, access to affordable and empathetic mental health support remains a

significant challenge, particularly in remote or low-resource regions. While state-of-the-art AI

systems like large language models offer advanced conversational capabilities, they often require

substantial computational resources, internet access, and data privacy safeguards—making them

less feasible for deployment in underserved areas. Interestingly, early AI systems like Eliza,

developed in the 1960s by Joseph Weizenbaum, used simple rule-based and pattern-matching

techniques to simulate human-like conversations in therapeutic settings. Though limited, Eliza’s

design demonstrates the enduring value of lightweight, scripted AI for focused emotional support

applications.

Imagine you're part of a development team in a non-profit organization focused on improving

mental health outreach in rural communities with limited internet access. Your team is tasked

with designing a lightweight, offline-capable chatbot inspired by Eliza’s conversational model.

The goal is not to replace licensed therapists, but to build a tool that can offer structured

emotional relief, active listening, and reflective questioning using ethical and non-intrusive

methods. This chatbot would be used to help individuals express their feelings and receive basic

mental wellness support without the need for deep learning models or large datasets.

This assignment challenges you to rethink and modernize Eliza’s architecture for current

needs while maintaining simplicity, explainability, and safety. You must consider technical

design, user interface, conversation scripts, and safeguards to prevent misuse or over-reliance.

The final system should align with Eliza’s core principles—recognizing keywords, responding

with empathy, and maintaining a non-judgmental tone—while adapting the experience to fit

modern psychological understanding and diverse user needs.

Answer the following Questions (Eliza-Focused)

1. How can Eliza’s rule-based architecture be redesigned as a mental health chatbot tailored

for underserved communities? What types of conversation scripts and pattern-matching

rules would make it both empathetic and safe?

Answer:

Eliza was a simple chatbot that replied to people by looking for keywords in their sentences. To modernize it:

• We can add more caring responses like:
  “It sounds like that’s been really hard for you. Do you want to talk more about it?”

• Set up rules to recognize emotional words like “sad,” “angry,” or “stressed,” and respond supportively.

• Make conversation scripts that feel natural and supportive, for example:

  • “How have you been feeling lately?”

  • “Sometimes just talking about it can really help.”

• Safety Rules:

  • Include escalation triggers for keywords indicating distress (e.g., “suicide”, “harm”) with local support referrals.

  • Limit session length and frequency to avoid dependency.

  • Avoid advice-giving; focus on listening and reflection.

2. What are the ethical, cultural, and psychological factors to consider when deploying a

basic therapeutic chatbot in real-world environments? How can the system ensure user

trust, emotional safety, and prevent dependency?

Ethical: Ensure transparency that it’s not a human or a therapist. Implement privacy by design with no data storage or transmission.

Cultural: Localize responses to respect language, values, and norms. Avoid western-centric emotional scripts.

Psychological: Focus on active listening, validation, and gentle redirection. Avoid making diagnoses.

Safety Mechanisms:

• Display disclaimers before each session.

• Include “exit” keywords and help resources.

• Use rotating conversational scripts to avoid repetitive or robotic tone.

3. Describe the technical framework of your Eliza-inspired system. How would you

implement it to run offline or on low-power devices? What technologies or platforms

would you choose, and why?

Languages/Tools: Python, SQLite for lightweight storage, or embedded systems like Raspberry Pi.

Architecture:

• Predefined rule-based engine using regex or decision trees.

• Local storage of conversation scripts in JSON format.

Platform: Android APK or desktop app for use on low-spec devices.

Offline Capability:

• No cloud dependencies.

• Scripts and responses stored locally.

• Text-based interface, optionally with voice input for accessibility.

4. Compare Eliza’s symbolic AI approach with modern conversational AI tools (e.g.,

ChatGPT, Replika). In which scenarios might Eliza-style systems still provide advantages

today, and how can symbolic reasoning be integrated with newer AI technologies?

Symbolic AI (Eliza):

• Pros: Transparent, low-resource, controllable.

• Cons: Rigid, lacks true understanding.

Modern AI (e.g., ChatGPT):

• Pros: More fluent, responsive, adaptive.

• Cons: Requires internet, opaque reasoning, higher risk of hallucinations.

Where Eliza-Style Wins:

• Low-bandwidth environments.

• High-control and explainability needed.

• Quick-to-deploy systems with limited resources.

Hybrid Possibility:

• Combine symbolic response templates with embedded NLP models for better parsing while maintaining control and simplicity.

Case Study 2:

Reimagining MACSYMA for Enhancing Digital Literacy Among Senior Citizens

In a world increasingly driven by digital interactions, senior citizens often face barriers when

accessing essential services such as online banking, healthcare portals, and digital

communication platforms. These challenges are compounded by technology anxiety, cognitive

load, and unfamiliarity with modern, often opaque, AI-driven interfaces. While current AI

systems offer advanced features, they typically rely on cloud-based infrastructure and neural

networks that lack explainability and are difficult for elderly users to trust or comprehend.

This case study invites you to revisit the Macsyma AI system, developed in the 1960s at MIT,

which specialized in symbolic mathematical reasoning and transparent problem-solving.

Rather than focusing on answers alone, Macsyma emphasized step-by-step explanation and

logical clarity—qualities that remain crucial for building accessible tools today. The goal is to

adapt Macsyma’s structured approach to develop a cognitive assistant for elderly users that can

help interpret complex personal information—such as billing statements, medication

schedules, insurance forms, and service guidelines—in a clear, understandable format.

Your task is to design a conceptual framework or prototype for a symbolic reasoning-based

assistant that empowers senior citizens to independently understand and navigate digital content.

The assistant should prioritize clarity, transparency, and simplicity over automation or

personalization. Unlike neural networks, the proposed system should operate on a set of human-

readable rules and logic, aligning with Macsyma's original architecture to deliver educational and

supportive explanations in user-friendly language.

Answer the following Questions (Macsyma-Focused)

1. How can Macsyma’s symbolic logic techniques be adapted into a user-friendly AI tool

that helps senior citizens understand complex documents or perform structured tasks?

What features would ensure interpretability and reduce user frustration?

What is Symbolic Logic?
It’s a rule-based system — it follows set steps to understand and explain information clearly.

How It Helps:

• Can read documents and pick out key info like:
  “What is the amount due?”
  “What is the last date to pay?”

• Can explain things slowly, like:
  “This extra fee is a late fine because you paid after the due date.”

Helpful Features for Seniors:

• Button to explain any word or term: “Explain this term”

• Step-by-step instructions:
  “First, let’s find the date. Now, let’s check the amount to pay.”

• Use icons, colors, and highlights to make things easier to understand

2. What user interface elements and interaction patterns should be integrated into the

assistant to make symbolic reasoning outputs digestible and helpful for senior citizens

with varying levels of tech literacy and cognitive ability?

User Interface Design:

• Large fonts, high-contrast text, simple navigation.

• Voice-over and text-to-speech options.

Interaction Patterns:

• Wizard-style tasks with simple questions.

• Undo/Redo and “Repeat last explanation” buttons.

• Allow switching between simplified and detailed views.

Cognitive Support:

• Use analogies (e.g., “Think of this like a library fine…”).

• Provide confirmation prompts to avoid confusion.

3. Propose a lightweight, offline-capable technical framework for your Macsyma-inspired

assistant. How would symbolic processing be implemented, and what constraints (e.g.,

language, device compatibility, privacy) need to be addressed?

Languages and Tools:

• Python (easy to use and lightweight)

• SymPy (for doing symbolic math or logic)

• Tkinter or Kivy (to build the app interface)

How It Works (Architecture):

• Uses built-in rules to read and explain documents

• Can highlight mistakes and guide users step-by-step

• Works in different languages (Urdu, Punjabi, etc.)

What It Needs:

• Works on older computers and tablets

• All files and rules are saved on the device — nothing is sent online

• Can store data safely (encrypted or deleted after use)

4. Compare the symbolic AI model used in Macsyma with modern deep learning-based

systems in terms of explainability, control, and ethical deployment. In what use cases

might symbolic AI offer a safer, more transparent solution for elderly users?

Symbolic AI (Macsyma):

• Pros: Transparent, predictable, user-explainable.

• Cons: Less flexible, requires more manual rule design.

Deep Learning Systems:

• Pros: Adaptive, natural language fluent.

• Cons: Black-box logic, may overwhelm or confuse users, requires training data.

When Symbolic Wins:

• Interpreting standardized documents (bills, medical forms).

• High-stakes tasks requiring clarity and auditability.

Ethical Edge:

• Symbolic systems offer control, ensure accuracy, and help reduce reliance on opaque automation.