Week10Lecture10_Production_Machine

Page 1: Lecture Access

• HOW TO FIND LECTURE 9 BILINGUALISM

  • Navigate: Canvas → Media Gallery → Click on the "Media" tab next to "Playlists"

Page 2: Final Exam Reminder

• FINAL EXAM NEXT WEEK (36% OF FINAL GRADE)

  • Date & Time: 7-10 pm on 12/13/2024 (FRIDAY)

  • Location: MOS 0114 (the same classroom as usual)

  • Format: In-person, 50 Multiple-choice questions

    • 25 questions from Week 0-6, 25 questions from Week 7-10

  • Materials: No need to bring your own scantron

  • OSD students: Schedule your exam ASAP

  • Additional office hours during Finals Week (refer to Canvas announcement)

Page 3: Final Grade Notification

• FINAL GRADE

  • Extra credit will not be added to your Canvas final grade until all other grades are finalized.

  • Hold off on emailing about your final grade until the announcement has been made on Canvas that grading is complete.

  • Post-final exam: Request double-checks on grading errors, but refrain from asking for adjustments for other reasons.

Page 4: Lecture Information

• Language Production

  • Date: Nov 5, 2024

  • Course: PSYC 145 Guest Lecture by Thomas Morton

Page 5: Self-introduction

• Content: "Who am I?"

Page 6: Outline of Lecture Topics

• Reasons to study language production

• Methods of studying language production

• Definition of speech error

• Insights from speech errors on production mechanisms

• Process of speaking: From intention to articulation

• Preparatory orientation before speaking

• Planning duration and its constraints

• Impact of past experiences on future production

• Word recall difficulties

• Nature of speech mistakes

• Variance in mistakes and error recognition

• Comprehensive models of speech production

Page 7: Complexity of Language

• Language is a complex, interconnected system:

  • Children acquire language systems naturally.

  • Comprehension involves active listening, prediction, analysis, and inference.

  • Structured representations and cognitive mechanisms are essential.

  • Learning includes insights gained from studying language use.

Page 8: Discovery in Language

• Quote: "Discovery is the ability to be puzzled by simple things"

  • Awareness of automatic processes in speaking.

  • Noam Chomsky's perspective on the simplicity of speaking despite its complexity.

Page 9: Language Production Mechanics

• Language production includes multiple levels:

  • Precise motor control necessary.

  • Progression: Sounds → Words → Sentences → Discourse and dialogue.

Page 10: Speaking Process Initiation

• Initiation in speaking: Begins with an idea or intention.

  • Planning starts at the highest intention level before motor control.

Page 11: Understudied Language Production

• Language production remains lesser-studied.

• Challenges in conducting research on sentence production versus understanding.

• Observation in speech is often easier than eliciting spoken responses.

Page 12: Time Allocation in Language Research

• Quote: "If Language Production is so important, why does it get 1 hour on the last day?"

  • Reflects the current challenges faced by researchers in language production.

Page 13: Early Research Methods in Language Production

• Naturalistic Research:

  • Early studies relied on observational data instead of experimental designs.

  • Examples include recording speech errors or conducting observational studies.

Page 14: Speech Errors in Linguistics

• Speaker: Victoria Fromkin (1973)

  • Errors can provide linguistic insight into speech production mechanisms.

  • Examples of speech errors: Spoonerisms, Freudian slips.

    • e.g., "light a fire" → "fight a liar"

Page 15: Reiteration of Speech Errors

• More examples of speech errors details:

  • Spoonerisms and implications for linguistic theory.

Page 16: Structure of Language Production

• Language production follows hierarchical processes to transform thoughts into fluent speech.

  • Active participation of speakers involves anticipating and correcting potential errors.

Page 17: Error Analysis in Language Production

• Types of speech errors are indicators of computational level:

  • Anticipation errors: e.g., "Baris is the most beautiful city"

  • Perseveration errors: e.g., "That kid was escorking us"

  • Exchange errors: e.g., "Seymour sliced the knife with a salami"

  • Substitution errors: e.g., "Till your heart’s dissent"

Page 18: Patterns from Speech Errors

• Identification of four error types:

  • Anticipation

  • Perseveration

  • Exchange

  • Substitution

Page 19: Error Patterns in Speech

• Analyzing clusters of errors can reveal various linguistic mechanisms at work.

Page 20: Insights from Speech Errors

• Speaker: Gary Dell (1989)

  • Errors suggest language production occurs at discrete levels of representation.

  • Speakers plan speech level-by-level rather than blending levels.

Page 21: Language Production Model

• Levelt’s Model of Language Production:

  1. Conceptualizer (intentions)

  2. Formulating messages with grammar and phonology

  3. Word retrieval from lexicon

  4. Monitoring of speech plans

  5. Translation into articulations

  6. Listening for errors post-speech

Page 22: Eye Fixation Studies

• Study by Zenzi Griffin and Kay Bock (2000)

  • Initial eye fixations reflect the order of spoken subjects.

Page 23: Structural Priming

• Research by Kay Bock (1986)

  • Grammar and sound rules shape messages.

  • Previous syntactic structures bias future speech output.

Page 24: Structural Alternation Examples

• Demonstration of active versus passive sentence structures and their effects on speech production.

Page 25: Priming Tasks

• Participants repeat a sentence prime before describing an image, showing priming effects.

Page 26: Results of Structural Priming Task

• Findings: Previous exposure to structures impacts present speech patterns.

Page 27: Contrast in Structural Priming Results

• Evidence of individuals favoring certain syntactic structures based on prior experience.

Page 28: Verb-Boost Effect in Structural Priming

• Research by Martin Pickering and Holly Branigan (1998)

  • Verb repetition strengthens the priming effect on sentence structures.

Page 29: Analysis of Verb-Boost Effect

• Hartsuiker et al. (2006) findings on word-boost effects in speech production.

Page 30: Long-term Effects of Verb-Boost

• Comparison of long-term structural changes versus shorter-term response patterns in speech production.

Page 31: Tip of the Tongue (TOT) States

• Situations where speakers know a word's meaning but can't access its phonology.

Page 32: Gendered Language and Inner Speech

• In gendered languages, individuals still guess gender accurately even when phonological recall fails.

Page 33: Inner Speech Research

• Studies by Gary Dell investigating how inner monitoring influences error detection.

Page 34: Inner Speech Theory

• Historical context from John B. Watson's experiments establishing the notion of inner speech.

Page 35: Error Reporting in Speech

• Investigating similarity in error detection between verbal and non-verbal tasks.

Page 36: Articulatory Errors in Similar Onsets

• Research indicates increased errors in tongue-twisters when attempting similar onset words.

Page 37: Impact of Articulatory Movement on Errors

• Studies showing that inner speech functionality impacts outer speech errors.

Page 38: Relationship Between Inner and Outer Speech

• Findings that articulatory movements mirror inner speech patterns regardless of articulation.

Page 39: Internal Monitoring Mechanisms

• Articulatory errors can manifest during inner speech, highlighting conceptual connections in speech production.

Page 40: Articulatory Planning Influence

• Evidence that physical articulatory errors arise from cognitive processes and planning inefficiencies.

Page 41: Monitoring Frequently Occurring Errors

• Slevc and Ferreira (2006) investigate sensitivity to semantic vs. phonological conflicts during speech.

Page 42: Stop-Signal Task

• Overview of the Stop-Signal Task: investigating participants' ability to halt speech under certain conditions.

Page 43: Measurement of Conflict Sensitivity

• Results demonstrate stronger responsiveness to phonological conflicts over semantic discrepancies.

Page 44: Sensitivity to Speech Errors

• Summary findings suggest heightened sensitivity to phonological over semantic errors.

Page 45: Role of Semantic Conflict in Monitoring

• Clarification that internal strategies focus on semantic errors while external monitoring deals with phonological output.

Page 46: Internal Speech Monitoring Model

• Revision of Levelt’s model emphasizes steps in message formulation until articulation.

Page 47: Recap on Language Complexity

• Language entangles various cognitive processes; successful production demands navigating those processes.

Page 48: Study Guide Components

• Summary of key concepts regarding language production principles and error handling.

Page 49: Language and Speech Production Insights

• Highlighting importance of visual orientation, past performance influences, and internal vs. external monitoring.

Page 50: Overview of Lecture on Machine and Language

• Lecture: IOB PSYC 145 - Fall 2024

Page 51: Extra Credit Protocol

• NO NEED TO REGISTER, not extra credit, only for polling

  • Raise hand to explain answers for extra credit

  • Must physically record your name for extra credit

Page 52: Lecture Agenda

• AGENDA

  • Topics to cover: AI Basics, Chatbots, Machine Learning, Ethics and Biases

Page 53: Artificial Intelligence Overview

• ARTIFICIAL INTELLIGENCE (AI)

Page 54: Views on Intelligence

• Different perspectives on intelligence:

  • General intelligence (g)

  • Separate intelligences

  • Hierarchical structures of tasks

Page 55: Turing Test Description

• MEASURING MACHINE INTELLIGENCE: THE TURING TEST

  • Definition: A human interrogator assesses machine intelligence by questioning both a machine and a human, aiming to discern who is who.

Page 56: Limitations of the Turing Test

• Limitations include various aspects of intelligence not covered such as non-verbal behaviors or non-factual systems.

Page 57: Interactive Human or Not? Game

• HUMAN OR NOT?

  • Public game assessing users' interactions with AI.

Page 58: AI Interaction Example

• Example dialogue illustrating AI's responses and user interactions.

Page 59: Game Outcomes

• Results on interactions confirm the AI nature.

Page 60: Game with User Queries

• Interaction samples showcasing user engagement challenges with distinguishing real human responses.

Page 61: Human Interaction Outcomes

• Demonstration of interaction where a human successfully engages.

Page 62: Introduction to ELIZA

• EARLIEST CHATBOT: ELIZA

  • Overview: Pattern-matching chatbot deploying rule-based action to mimic therapist interactions.

Page 63: ELIZA in Action

• Demonstration of the ELIZA chatbot's capabilities in engaging users through predefined psychological responses.

Page 64: Modern Chatbot Compared to ELIZA

• Contrasting responses of ELIZA (1960s) with ChatGPT (modern AI) in handling user inquiries.

Page 65: Chatbot Feature Comparison

• Comprehensive comparison of rule-based chatbots versus Large Language Models (LLMs) with examples.

Page 66: Turing Test Results

• Findings on performances in the Turing Test, indicating advancements and comparative benchmarks against human intelligence.

Page 67: Machine Learning Explained

• HOW DO MACHINES "LEARN"?

Page 68: Types of Machine Learning

• Detailed insights into supervised, unsupervised, and reinforcement learning mechanisms.

Page 69: Reinforcement Learning Example Query

• A quiz question about reinforcement learning scenarios in animals and humans.

Page 70: Training Mechanism of LLMs

• Explanation of how LLMs utilize probability distributions to predict language sequences during training.

Page 71: ChatGPT Training Process

• Steps involved in training ChatGPT including reward model fundamentals and supervised policies.

Page 72: Language Acquisition Theories

• Revisiting language acquisition theories contrasting machine and human approaches in learning.

Page 73: Environmental Impact of LLMs

• Examination of carbon emissions generated by various machine learning models and their cumulative effects on the planet.

Page 74: Ethics and Biases in AI

• Overview of ethical considerations in designing AI systems.

Page 75: AI Biases

• Discussion on biases related to training data and design processes affecting ethical AI outcomes.

Page 76: Case Study on Racial Biases in AI

• Investigative report on racial biases in skin cancer detection AI systems and their implications.

Page 77: Gender Bias in AI Responses

• Study investigating gender biases in sentence attribution within AI models.

Page 78: AI Bias Correction Challenges

• Overview of the complexity involved in correcting biases inherent to AI systems and training data.

Page 79: Study Guide Focus

• Comprehensive review suggestions for the material covered in the lecture for exams.

Page 80: Key Questions for Examination

• Topics include the Turing Test, chatbot functionality, learning methods, ethical AI considerations.