Cognitive Computing and Cognitive Models

Cognitive Computing:
  - A branch of artificial intelligence (AI) aimed at simulating human thought processes in computerized models.
  - Mimics human thinking, reasoning, and learning from experience.
  - Utilizes technologies such as:
    - Machine learning
    - Natural language processing (NLP)
    - Speech recognition
    - Image processing
    - Neural networks

Adaptive: Learns and adapts from data patterns and feedback.
Interactive: Communicates naturally with users.
Contextual: Understands meaning based on context (e.g., time, location, domain).
Iterative: Improves performance with repeated interaction.
Evidence-based: Provides reasoning and justification for recommendations.

Goal:
- Cognitive Computing: Augment human intelligence.
- Traditional AI: Automate decision-making or tasks.
Approach:
- Cognitive Computing: Mimics human cognition and reasoning.
- Traditional AI: Executes pre-defined rules and logic.
Interaction:
- Cognitive Computing: Human-like, natural language interaction.
- Traditional AI: Often limited to command-based or structured input.
Learning:
- Cognitive Computing: Learns continuously from new data and feedback.
- Traditional AI: Often relies on initial training or rule-based programming.
Handling of Uncertainty:
- Cognitive Computing: Can deal with ambiguity, incomplete, or unstructured data.
- Traditional AI: Requires structured, complete data for accurate processing.
Examples:
- Cognitive Computing: IBM Watson, Google DeepMind, Microsoft Cognitive Services.
- Traditional AI: Expert systems, traditional rule-based chatbots, automation bots.
Use Case:
- Cognitive Computing: Supports complex decision-making (e.g., healthcare diagnosis).
- Traditional AI: Automates well-defined, repetitive tasks (e.g., sorting emails).

Definition: IBM Watson is a powerful cognitive computing system developed by IBM to understand, reason, learn, and interact using natural language and machine learning technologies.
Historical Context: Named after IBM's first CEO, Thomas J. Watson, it gained fame in 2011 by winning the quiz show "Jeopardy!" defeating human champions.

Natural Language Processing (NLP):
- Understands and interprets human language (both spoken and written).
- Capable of answering questions, summarizing texts, and extracting meaning from documents.
Machine Learning:
- Learns continuously from new data to improve accuracy and relevance.
- Supports adaptive learning models.
Data Analysis:
- Processes both structured and unstructured data (text, audio, images).
- Identifies patterns, trends, and insights in vast data sets.
Evidence-Based Reasoning:
- Provides reasoning behind conclusions and recommendations.

IBM Watson exemplifies the principles of cognitive computing, enhancing the capabilities and effectiveness of human decision-making by providing data-driven insights.

Healthcare:
- Watson for Oncology assists doctors by suggesting treatment options based on patient data and medical literature.
- Assists in drug discovery and clinical trial matching.
Finance:
- Analyzes risks for banks, detects fraud, and personalizes financial advice.
Customer Service:
- Powers chatbots and virtual agents to handle customer queries effectively.
Education:
- Provides personalized learning experiences and tutoring systems for students.
Legal and Compliance:
- Analyzes legal documents, supports due diligence, and identifies compliance risks.

Definition:
- These models provide answers with supporting evidence and confidence levels, aiding informed decision-making.
Characteristics:
- Represent "what" a person knows—facts and information that can be explicitly recalled and verbalized.
Examples:
- Using declarative memory to remember a formula (e.g., $F = ma$ ).
Applications:
- Educational software modeling how students learn facts.
- Simulations of retrieval of facts during decision-making processes.

Definition:
- Use formal logic systems to simulate human reasoning, capturing how conclusions are derived from known premises.
Key Concepts:
- Support deductive and sometimes inductive reasoning.
Examples:
- Expert systems such as MYCIN for medical diagnosis.
- Cognitive modeling in tasks like syllogisms or conditional logic.

Focus:
- Knowledge representation (what is known) for declarative models.
- Logical reasoning (how to reason) for logic-based models.
Knowledge Type:
- Declarative: factual, episodic, semantic.
- Logic-Based: rules, inferences, propositions.
Nature:
- Declarative: Memory-based.
- Logic-Based: Rule-based and logic-driven.
Use in AI:
- Declarative: Learning systems, cognitive tutors.
- Logic-Based: Expert systems, rule-based agents.

Definition:
- Use probabilistic reasoning to infer and update beliefs based on uncertain or incomplete information.
Core Idea:
- Based on Bayes' Theorem, which updates the probability of a hypothesis given new evidence.

Applications:
- Used in understanding human reasoning under uncertainty and decision-making practices.

Definition:
- Simulate cognition using networks of simple, interconnected units resembling neurons.
Core Idea:
- Knowledge is stored as patterns of activation across the network instead of explicit rules.
Key Characteristics:
- Emphasizes parallel processing and learns from examples through weight adjustment (e.g., backpropagation).
Applications:
- Effective in visual perception, language understanding, memory, and cognitive development.

Definition:
- Refers to the use of technology to enhance human intelligence rather than replace it.
Key Features:
  - Human-centered: Keeps humans in the loop.
  - Assistive, not autonomous: Aids informed decision-making without taking control.
  - Context-aware: Understands user environment and intent.
  - Explainable: Offers reasons for predictions and recommendations.

Healthcare: Supports diagnostic processes and personalized treatment planning.
Finance: Enhances risk analysis and client advisory services.
Customer Service: Powers chatbots and virtual agents to improve service quality.
Education: Provides tailored learning aids and feedback systems.
Business Intelligence: Assists in data-driven decision-making through simulations and analyses.

Collaboration: The premise that "Humans + Machines = Better decisions" emphasizes the collaborative nature of augmented intelligence, fostering improved outcomes in decision-making scenarios.