Defensive Avoidance in Paranoid Delusions: Experimental and Compu- Tational Approaches
Total Page:16
File Type:pdf, Size:1020Kb
DEFENSIVE AVOIDANCE IN PARANOID DELUSIONS: EXPERIMENTAL AND COMPUTATIONAL APPROACHES A THESIS SUBMITTED TO THE UNIVERSITY OF MANCHESTER FOR THE DEGREE OF DOCTOR OF PHILOSOPHY (PHD) IN THE FACULTY OF MEDICAL AND HUMAN SCIENCES MICHAEL MOUTOUSSIS SCHOOL OF PSYCHOLOGICAL SCIENCES 2011 2 CONTENTS Abbreviations & Key Symbols 11 Abstract 15 Declaration 17 Copyright 19 Submission in Alternative Format 21 Acknowledgements 23 The Author 25 1 Introduction 27 1.1 Importance................................. 27 1.2 Schools of thought on paranoid delusions................. 29 1.2.1 Paranoid delusions in psychiatry.................. 29 1.3 The psychological understanding of paranoia............... 30 1.3.1 Psychodynamics of paranoia.................... 30 1.3.2 Cognitive-behavioural approaches to paranoia........... 34 1.4 Thesis outline................................ 36 2 Paranoia and Conditioned Avoidance 37 2.1 Summary.................................. 37 2.2 The importance of threat-related content in delusions........... 38 2.3 Threats & Attribution theory........................ 39 2.3.1 ‘Poor-me’ and ‘bad-me’ paranoia................. 41 2.4 The Conditioned Avoidance paradigm................... 42 2.4.1 Linking paranoid delusions and the CAR............. 43 2.4.2 The experience of threat in the CAR and in paranoia....... 44 2.4.3 Safety behaviours may help maintain paranoia.......... 44 2.4.4 The avoidance of internal states in Poor-me paranoia....... 45 3 CONTENTS 2.5 CAR and paranoia: Conclusions...................... 46 3 Computational methods for cost-based decision making 49 3.1 Summary.................................. 49 3.2 A modelling framework: ‘Cached’ & ‘tree-search’ models........ 50 3.3 Temporal-difference models........................ 52 3.3.1 Fixed parameters used in the Advantage-learning model of avoidance......................... 56 3.4 Tree-search & related models........................ 57 3.4.1 Costed Bayesian model....................... 57 3.4.2 Sequential probability ratio test model............... 59 3.5 Fitting models to data........................... 60 3.5.1 Full expectation-maximisation fitting............... 61 3.5.2 Experimental Bayesian distribution................ 65 3.6 Model Evaluation.............................. 65 3.6.1 Bootstrap methods......................... 66 3.6.2 Model comparison and the Bayesian information criterion.... 71 3.7 Computational methods: Conclusions................... 71 4 A Temporal Difference Account of Avoidance 73 4.1 Summary.................................. 73 4.2 Conditioned avoidance (CAR) and reinforcement learning........ 74 4.2.1 The CAR experimental paradigm................. 75 4.3 The Advantage-learning model....................... 78 4.4 Results.................................... 80 4.4.1 Simulation of escape-from-fear learning.............. 80 4.4.2 Simulation of normal CAR learning................ 84 4.4.3 Simulation of dopaminergic manipulations in the CAR...... 86 4.5 Discussion.................................. 86 4.5.1 Neurobiological substrate..................... 90 4.5.2 Relevance of the Avoidance model................. 92 4.6 Temporal-difference modelling of avoidance: Conclusions................................. 94 5 Avoidance and Jumping-to-Conclusions in Paranoia 95 5.1 Summary.................................. 95 5.2 Jumping-to-Conclusions: an introduction................. 96 5.2.1 Delusions & Bayesian reasoning.................. 96 5.2.2 Explanations of the Jumping-to-Conclusions (JTC) bias..... 97 5.2.3 The ideal-observer Bayesian approach............... 101 5.2.4 Comparison with the Sequential Probability Ratio Test...... 103 4 CONTENTS 5.3 Methods................................... 105 5.4 Results.................................... 106 5.5 Discussion.................................. 111 5.5.1 High-noise processing vs. the high-sampling-cost hypothesis... 111 5.5.2 Paranoid decision-making and noise................ 112 5.5.3 Bayesian vs. threshold-driven decisions in paranoia........ 113 5.5.4 Methodological advances...................... 114 5.5.5 Limitations of the present study.................. 116 5.6 Modelling the JTC bias: Conclusions................... 117 6 Empirical study of defensive avoidance 119 6.1 Summary.................................. 119 6.2 Background & importance......................... 120 6.3 Ethical issues................................ 122 6.3.1 The possibility of distressing or harming participants....... 122 6.3.2 Vulnerable groups & informed consent issues........... 123 6.4 Empirical study: Design & research Instruments.............. 123 6.4.1 Participant groups......................... 123 6.4.2 Inclusion & exclusion criteria................... 124 6.4.3 Data-gathering procedures and instruments............ 125 6.4.4 Outcome measures & power calculations............. 129 6.4.5 Analysis strategy.......................... 130 6.5 Interview protocol.............................. 130 6.5.1 Identification & recruitment of participants............ 130 6.5.2 The research interview....................... 131 6.6 Results................................... 131 6.6.1 Summary.............................. 131 6.6.2 Participant welfare......................... 132 6.6.3 Key hypotheses........................... 133 6.6.4 Regression analyses........................ 146 6.7 Discussion.................................. 153 6.7.1 Poor-me paranoia is not due to increased defensive avoidance.. 153 6.7.2 Interpretation of the results..................... 153 6.7.3 Limitations of the present empirical study............. 158 6.7.4 Future directions.......................... 158 6.8 Empirical investigation of defensive avoidance: Conclusions..................... 159 7 Defensive avoidance in paranoia: Synthesis and general discussion 161 7.1 Summary.................................. 161 5 CONTENTS 7.2 An integrative approach to paranoid delusions............... 162 7.2.1 Avoidance and self-discrepancies in paranoia........... 163 7.2.2 The role of self-esteem in paranoia................. 167 7.3 Limitations................................. 168 7.4 Implications for future research & therapy................. 170 7.4.1 Clinical implications........................ 170 7.4.2 Research implications....................... 171 Contribution 175 References 177 Final word count: 52927 6 LIST OF FIGURES 1.1 Thoughts and feelings related to paranoia are common and distressing.. 28 3.1 State space for the TD model of the CAR................. 53 3.2 Probability to decide at each step of the beads task (CB model)...... 62 3.3 How typical an example of model output would the data be?....... 67 3.4 Monte-Carlo simulations of experimental summary statistics....... 69 3.5 Monte-Carlo simulations of model macroparameters........... 70 4.1 Conditioned avoidance: Experimental interventions & response...... 75 4.2 CAR: Key experimental results....................... 77 4.3 Information flow in the actor-critic model................. 80 4.4 Escape-From-Fear simulation results.................... 82 4.5 Escape-From-Fear – Value Learning.................... 83 4.6 Escape-From-Fear – Advantage Learning................. 83 4.7 Simulation of normal CAR learning.................... 85 4.8 CAR learning under DA block....................... 87 4.9 CAR learning under reversal of DA block................. 88 5.1 Workings of the SPRT model........................ 104 5.2 Best-fit parameters and BIC values for different participant groups.... 109 5.3 Illustration of the effect of noise in the healthy and paranoid case..... 110 5.4 Differences between Costed-Bayesian and SPRT models revealed by ur- gency plot.................................. 115 5.5 Marginal posterior density plot reveals possible fine structure of data... 116 6.1 Request for ’Feared Self’ attributes..................... 127 6.2 Persecution and deservedness according to clinical group......... 134 6.3 Self-Discrepancies and paranoia...................... 138 6.4 Depression and self-discrepancies..................... 140 6.5 AAQ-2 and MC scores in paranoia..................... 142 6.6 Distribution of change in response time with IA discrepancy....... 144 7 8 LIST OF TABLES 2.1 Positive symptoms of patients recruited to the SoCRATES study..... 39 5.1 Best-fit parameters and BIC values for JTC tasks............. 107 6.1 Key descriptive statistics.......................... 135 6.2 Paranoia grouping vs. Self-Discrepancies................. 137 6.3 Paranoia grouping vs. AAQ-2 and social desirability........... 141 6.4 Depression × paranoia analysis....................... 141 6.5 Interactive measures – descriptive statistics................ 145 6.6 Overall change in median response times................. 145 6.7 ANOVA of response times......................... 145 6.8 ANOVA of Fraction of high-engagement responses............ 146 6.9 Regression analysis for auxiliary variables................. 149 6.10 Regression analysis - PADS-P....................... 149 6.11 Regression analysis - undeserved paranoia................. 150 6.12 Exploratory regression analysis - PADS-P................. 152 6.13 Exploratory regression analysis - AAQ-2.................. 152 9 10 ABBREVIATIONS & KEY SYMBOLS Abbreviations 5HT 5 hydroxytryptamine (Serotonin; if followed by additional symbols, e.g. 5HT2a, it denotes a serotonin receptor type), 43 AAQ-2 ‘Acceptance and action questionnaire’, second version, 131 ACT Acceptance and Commitment Therapy, 162 AID Actual–Ideal discrepancy, 121 ANOVA Analysis of variance, 130 AR Avoidance