Robust Intelligence and Trust in Autonomous Systems

Preface

AAAI-2014 Spring Symposium Organizers

AAAI-2014 Spring Symposium: Keynote Speakers

Symposium Program Committee

Contents

1 Introduction

1.1 The Intersection of Robust Intelligence (RI) and Trust in Autonomous Systems

1.2 Background of the 2014 Symposium

1.3 Contributed Chapters

References

2 Towards Modeling the Behavior of Autonomous Systems and Humans for Trusted Operations

2.1 Introduction

2.2 Understanding the Value of Context

2.3 Context and the Complexity of Anomaly Detection

2.3.1 Manifolds for Anomaly Detection

2.4 Reinforcement Learning for Anomaly Detection

2.4.1 Reinforcement Learning

2.4.2 Supervised Autonomy

2.4.3 Feature Identification and Selection

2.4.4 Approximation Error for Alarming and Analysis

2.4.5 Illustration

2.4.5.1 Synthetic Domain

2.4.5.2 Real-World Domain

2.5 Predictive and Prescriptive Analytics

2.6 Capturing User Interactions and Inference

2.7 Challenges and Opportunities

2.8 Summary

References

3 Learning Trustworthy Behaviors Using an Inverse Trust Metric

3.1 Introduction

3.2 Related Work

3.2.1 Human-Robot Trust

3.2.2 Behavior Adaptation

3.3 Agent Behavior

3.4 Inverse Trust Estimate

3.5 Trust-Guided Behavior Adaptation

3.5.1 Evaluated Behaviors

3.5.2 Behavior Adaptation

3.6 Evaluation

3.6.1 eBotworks Simulator

3.6.2 Experimental Conditions

3.6.3 Evaluation Scenarios

3.6.3.1 Movement Scenario

3.6.3.2 Patrolling Scenario

3.6.4 Trustworthy Behaviors

3.6.5 Efficiency

3.6.6 Discussion

3.7 Conclusions

References

4 The “Trust V”: Building and Measuring Trust in Autonomous Systems

4.1 Introduction

4.2 Autonomy, Automation, and Trust

4.3 Dimensions of Trust

4.3.1 Trust Dimensions Arising from Automated Systems Attributes

4.3.2 Trust Dimensions Arising from Autonomous Systems Attributes

4.3.3 Another Trust Dimension: SoS

4.4 Creating Trust

4.4.1 Building Trust In

4.5 The Systems Engineering V-Model

4.6 The Trust V-Model

4.6.1 The Trust V Representation: Graphic

4.6.2 The Trust V Representation: Array

4.6.3 Trust V “Toolbox”

4.7 Specific Trust Example: Chatter

4.8 Measures of Effectiveness

4.9 Conclusions and Next Steps

A.1 Appendix

References10

5 Big Data Analytic Paradigms: From Principle Component Analysis to Deep Learning

5.1 Introduction

5.2 Wind Data Description

5.3 Wind Power Forecasting Via Nonparametric Models

5.3.1 Advanced Neural Network Architectures Application

5.3.2 Wind Speed Results

5.4 Introduction to Deep Architectures

5.4.1 Training Deep Architectures

5.4.2 Training Restricted Boltzmann Machines

5.4.3 Training Autoencoders

5.5 Conclusions

References

6 Artificial Brain Systems Based on Neural Network Discrete Chaotic Dynamics. Toward the Development of Conscious and Rational Robots

6.1 Introduction

6.2 Background

6.3 Numerical Simulations

6.4 Conclusion

References

7 Modeling and Control of Trust in Human-Robot Collaborative Manufacturing

7.1 Introduction

7.2 Trust Model

7.2.1 Time-Series Trust Model for Dynamic HRC Manufacturing

7.2.2 Robot Performance Model

7.2.3 Human Performance Model

7.3 Neural Network Based Robust Intelligent Controller

7.4 Control Approaches: Intersection of Trust and Robust Intelligence

7.4.1 Manual Mode

7.4.2 Autonomous Mode

7.4.3 Collaborative Mode

7.5 Simulation

7.5.1 Manual Mode

7.5.2 Autonomous Mode

7.5.3 Collaborative Mode

7.5.4 Comparison of Control Schemes

7.6 Experimental Validation

7.6.1 Experimental Test Bed

7.6.2 Experimental Design

7.6.2.1 Experiment Scenario

7.6.2.2 Controlled Behavioral Study

7.6.2.3 Imposing Fatigue

7.6.2.4 Experiment Procedure

7.6.2.5 Measurements and Scales

7.6.3 Experimental Results

7.6.3.1 Trust Model Identification Procedure

7.6.3.2 Manual Mode

7.6.3.3 Autonomous Mode

7.6.3.4 Collaborative Mode

7.6.4 Comparison and Conclusion

7.7 Conclusion

References

8 Investigating Human-Robot Trust in Emergency Scenarios: Methodological Lessons Learned

8.1 Introduction

8.2 Conceptualizing Trust

8.2.1 Conditions for Situational Trust

8.3 Related Work on Trust and Robots

8.4 Crowdsourced Narratives in Trust Research

8.4.1 Iterative Development of Narrative Phrasing

8.5 Crowdsourced Robot Evacuation

8.5.1 Single Round Experimental Setup

8.5.2 Multi-Round Experimental Setup

8.5.3 Asking About Trust

8.5.4 Measuring Trust

8.5.5 Incentives to Participants

8.5.6 Communicating Failed Robot Behavior

8.6 Conclusion

References

9 Designing for Robust and Effective Teamwork in Human-Agent Teams

9.1 Introduction

9.2 Related Work

9.2.1 Team Structure

9.2.2 Shared Mental Model and Team Situation Awareness

9.2.3 Communication

9.3 Experiment 1: Team Structure and Robustness

9.3.1 Testbed

9.3.2 Experiment Design

9.3.3 Results

9.3.3.1 Duplicated Work

9.3.3.2 Under Utilization of Vehicles

9.3.3.3 Infrequent Communication

9.4 Experiment 2: Information-Sharing

9.4.1 Independent Variables

9.4.2 Dependent Variables

9.4.3 Participants

9.4.4 Procedure

9.4.5 Results

9.4.5.1 Team Performance

9.4.5.2 Team Coordination

9.4.5.3 Workload

9.4.5.4 User Preference and Comments

9.5 Discussion

9.6 Conclusion

References

10 Measuring Trust in Human Robot Interactions: Development of the “Trust Perception Scale-HRI”

10.1 Introduction

10.2 Creation of an Item Pool

10.3 Initial Item Pool Reduction

10.3.1 Experimental Method

10.3.2 Experimental Results

10.3.3 Key Findings and Changes

10.4 Content Validation

10.4.1 Experimental Method

10.4.2 Experimental Results

10.5 Task-Based Validity Testing: Does the Score Change Over Time with an Intervention?

10.5.1 Experimental Method

10.5.2 Experimental Results

10.5.2.1 Individual Item Analysis

10.5.2.2 Trust Score Validation

10.5.2.3 40 Items Versus 14 Items

10.6 Task-Based Validity Testing: Does the Scale Measure Trust?

10.6.1 Experimental Method

10.6.2 Experimental Results

10.6.2.1 Correlation Analysis of the Three Scales

10.6.2.2 Pre-post Interaction Analysis

10.6.2.3 Differences Across Scales and Conditions

10.6.3 Experimental Discussion

10.7 Conclusion

10.7.1 The Trust Perception Scale-HRI

10.7.2 Instruction for Use

10.7.3 Current and Future Applications

References

11 Methods for Developing Trust Models for Intelligent Systems

11.1 Introduction

11.2 Prior Work in the Development of Trust Models

11.2.1 Trust Models

11.2.2 Trust in Human-Robot Interaction (HRI)

11.3 The Use of Surveys as a Method for Developing Trust Models

11.3.1 Methodology

11.3.2 Results and Discussion

11.3.3 Modeling Trust

11.4 Robot Studies as a Method for Developing Trust Models

11.4.1 Methodology

11.4.2 Results and Discussion

11.4.2.1 Reducing Situation Awareness (SA)

11.4.2.2 Providing Feedback

11.4.2.3 Reducing Task Difficulty

11.4.2.4 Long-Term Interaction

11.4.2.5 Impact of Timing of Periods of Low Reliability

11.4.2.6 Impact of Age

11.4.3 Modeling Trust

11.5 Conclusions and Future Work

References

12 The Intersection of Robust Intelligence and Trust: Hybrid Teams, Firms and Systems

12.1 Introduction

12.1.1 Background

12.2 Theory

12.3 Outline of the Mathematics

12.3.1 Field Model

12.3.2 Interdependence

12.3.3 Incompleteness and Uncertainty

12.4 Evidence of Incompleteness for Groups

12.4.1 The Evidence from Studies of Organizations

12.4.2 Modeling Competing Groups with Limit Cycles

12.5 Gaps

12.6 Conclusions

References