Emotional Understanding The Adaptability of Accurate Emotional Predictions DISSERTATION zur Erlangung des akademischen Grades doctor rerum naturalium (Dr. rer. nat.) im Fach Psychologie eingereicht an der Mathematisch-Naturwissenschaftlichen Fakultät II Humboldt-Universität zu Berlin von Frau Dipl.-Psych. Michaela Turß Präsident der Humboldt-Universität zu Berlin: Prof. Dr. Jan-Hendrik Olbertz Dekan der Mathematisch-Naturwissenschaftlichen Fakultät II: Prof. Dr. Elmar Kulke Gutachter: 1. Prof. Dr. Wolfgang Scholl 2. Prof. Dr. Ralf Schulze 3. Prof. Gerald Matthews, PhD eingereicht am: 06. März 2013 Tag der mündlichen Prüfung: 13. September 2013 Abstract In the ability model of emotional intelligence by Mayer and Salovey (1997), emotional understanding is a prerequisite for emotion regulation. Knowing which emotions occur in which situations should be beneficial and adaptive. One of the subtests for emotional understanding asks for likely emotional reactions in hypothetical situations. In contrast, Gilbert and Wilson (2003) argue that characteristic biases in affective forecasting are adaptive. The current thesis aims to measure accuracy of emotional predictions in a na- tural setting and examines its adaptive value. In the anxiety study, public officialswereaskedtopredictfutureemotionsinanimportanttest(N=143). The second study focused on freshman student work-groups (N=180 in 43 groups). Group members predicted interpersonal feelings for each other (af- fection, satisfaction with the collaboration, fun, and anger). In both studies, accuracy of emotional predictions is defined as low bias (i.e. Euclidean di- stance) and high correspondence (i.e. profile correlation). The round robin design in the work-group study also allows to decompose accuracy following Cronbach (1955). In both studies, alowbias was adaptive in terms ofstrong criteria, also incrementally over and above intelligence and personality alo- ne.Accuracywaspartlyrelatedtogeneralknowledgebutnottointelligence. Associations to emotional intelligence were inconsistent. Accuracy as corre- spondence is theoretically interesting but much less reliable. There is some evidence for its adaptive value on a group level but no indication of in- cremental validity. Future research should focus on specific situations and specific emotions. Also, processes underlying affective forecasts should be evaluated in detail. i Zusammenfassung Im Rahmen des Leistungsansatzes von emotionaler Intelligenz sehen Mayer und Salovey (1997) Emotionsverstaendnis als Voraussetzung für Emotions- regulation. Es sollte nützlich sein zu wissen, wie man sich in bestimmten Situationen fühlen wird. Zur Messung werden unter anderem Vignetten eingesetzt, in denen Emotionen für hypothetische Situationen vorhergesagt werden. Im Gegensatz dazu postulieren Gilbert und Wilson (2003) charak- teristische Fehler bei affektiven Vorhersagen, die motivational günstig sind. In der vorliegenden Arbeit wird die Akkuratheit emotionaler Vorhersagen im natürlichen Umfeld untersucht, um dessen adaptiven Wert zu beurtei- len. Zunächst sollten Beamtenanwärter ihre Emotionen in einer bedeuten- denTestsituationvorhersagen(N=143).DannwurdenstudentischeArbeits- gruppen(180Mitgliederin43Gruppen)gebeten,GefühlezwischendenMit- gliedern zu prognostizieren (Zuneigung, Zufriedenheit mit der Zusammen- arbeit, Freude und Ärger). Akkuratheit wurde als geringer Bias (euklidische Distanz) und hohe Korrespondenz (Profilkorrelation) definiert. Das Round Robin Design der zweiten Studie ermöglichte die Varianzzerlegung der Ak- kuratheit nach Cronbach(1955). In beiden Studien ist ein niedriger Bias ad- aptivinHinblickaufharteKriterien,auchinkrementellüberIntelligenzund Persönlichkeit hinaus. Bias hing teilweise mit Allgemeinwissen zusammen, abernichtmitIntelligenz.ZusammenhängezuemotionalerIntelligenzwaren inkonsistent. Die Akkuratheit als Korrespondenz ist theoretisch interessant aberdeutlichwenigerreliabel.AufGruppenebenekonntedieKorrespondenz Kriterien vorhersagen, aber es zeigte sich keine inkrementelle Validität. Zu- künftige Forschung sollte sich auf spezifische Situationen und spezifische Emotionen konzentrieren sowie die Prozesse untersuchen, die emotionalen Vorhersagen zugrunde liegen. ii Contents 1 General Introduction 1 2 Theoretical Background 4 2.1 Emotional Intelligence . . . . . . . . . . . . . . . . . . . . . . 4 2.2 Research on Emotional Predictions . . . . . . . . . . . . . . . 9 2.3 Accuracy Research . . . . . . . . . . . . . . . . . . . . . . . . 13 2.4 Research Questions . . . . . . . . . . . . . . . . . . . . . . . . 16 3 Study 1: Anxiety Study 19 3.1 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 3.1.1 Sample . . . . . . . . . . . . . . . . . . . . . . . . . . 19 3.1.2 Procedure . . . . . . . . . . . . . . . . . . . . . . . . . 20 3.1.3 Material . . . . . . . . . . . . . . . . . . . . . . . . . . 20 3.2 Results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 3.2.1 Descriptive Statistics . . . . . . . . . . . . . . . . . . . 24 3.2.2 Forecasts and Experiences over Time . . . . . . . . . . 25 3.2.3 Accuracy as Bias . . . . . . . . . . . . . . . . . . . . . 27 3.2.4 Accuracy as Correspondence . . . . . . . . . . . . . . 31 3.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 3.3.1 Bias, Correspondence, and other Abilities and Traits . 40 3.3.2 The Adaptability of Bias and Correspondence . . . . . 41 3.3.3 Connection to the Affective Forecasting Paradigm . . 42 iii 4 Study 2: Work-Group Study 44 4.1 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 4.1.1 Sample . . . . . . . . . . . . . . . . . . . . . . . . . . 44 4.1.2 Design and Procedure . . . . . . . . . . . . . . . . . . 45 4.1.3 Material . . . . . . . . . . . . . . . . . . . . . . . . . . 45 4.2 Results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 4.2.1 The Group as Unit of Analysis . . . . . . . . . . . . . 49 4.2.1.1 Descriptive Statistics . . . . . . . . . . . . . 49 4.2.1.2 Bias and Correspondence . . . . . . . . . . . 50 4.2.1.3 Accuracy Decomposition . . . . . . . . . . . 54 4.2.2 The Person as Unit of Analysis . . . . . . . . . . . . . 56 4.2.2.1 Perceiver and Target Effects . . . . . . . . . 56 4.2.2.2 Bias and Correspondence . . . . . . . . . . . 57 4.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 4.3.1 Stability of Bias and Correspondence . . . . . . . . . . 63 4.3.2 Bias, Correspondence, and other Abilities and Traits . 64 4.3.3 The Adaptability of Bias and Correspondence . . . . . 65 5 General Discussion 66 5.1 Theoretical Considerations. . . . . . . . . . . . . . . . . . . . 66 5.2 Strengths and Limitations . . . . . . . . . . . . . . . . . . . . 68 5.3 Implications for Future Research . . . . . . . . . . . . . . . . 69 5.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 References 73 iv List of Figures 3.1 Course of Predictions and Experiences (N = 128 - 141) . . . 27 3.2 AnxietyasaModeratoroftheRelationshipbetweenBiasand Coping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 3.3 Relationship between Emotional Predictions (x) and Expe- riences (y) for High and Low Anxiety, and High and Low Positive Coping . . . . . . . . . . . . . . . . . . . . . . . . . . 39 v List of Tables 3.1 Anxiety Study: Course of the Data Collection . . . . . . . . . 21 3.2 Sample Items for Positive and Negative Coping . . . . . . . . 24 3.3 Intercorrelations and Descriptive Statistics . . . . . . . . . . . 26 3.4 Repeated-measures ANOVA and Contrasts with Positive and Negative Affect as Outcome Variables . . . . . . . . . . . . . 28 3.5 First-order Correlations of Difference Scores and Criteria. . . 30 3.6 One-way ANOVA with Random Effects. Unrestricted Model 33 3.7 Estimating Experiences with Predictions . . . . . . . . . . . . 35 3.8 Moderators of Correspondence . . . . . . . . . . . . . . . . . 36 3.9 Adaptability of Correspondence . . . . . . . . . . . . . . . . . 38 4.1 IntercorrelationsandDescriptiveStatistics(GroupLevel,N=43) 51 4.2 Retest Reliability and Validity of Accuracy Measures (Group Level, N=43) . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 4.3 Validity of Accuracy Components (Group Level, N=43) . . . 55 4.4 EffectsforPerceivers,Targets,andTargetsoftheSecondDegree 58 4.5 Hierarchical Random Intercept Model for Task and Relation- ship Conflict . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 4.6 Hierarchical Random Intercept Model for Competence and Friendship . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62 vi Chapter 1 General Introduction In recent years, the concept of emotional intelligence (EI) has been recei- ving increasing attention. Publications on EI have grown exponentially over at least two decades, and economic interest creates considerable market demands regarding measurement and training. Whereas the concept is em- braced in public, it is controversially debated, if not severely criticized in academia. Itisgenerallyplausiblethatpeopledifferintheirabilityregarding emotions, i.e. that some are more likely to succeed in emotionally challeng- ing situations. Also, it is easy to appreciate the fact that emotion-related skills and characteristics, like dealing with stress, managing conflict or stay- ing optimistic, are valuable in life and many people may perceive someone to be emotionally gifted. EI proves useful to talk about success in everyday life, especially in social life. Nearly every useful addition to general intelli- gence can be projected into this concept. Also, it emphasizes strengths that go beyond scholastic achievement. When it comes to the theoretical basis, though, and to generally accepted definitions and measurement procedures, academic intelligence outclasses EI by far. The current thesis aims to add to the comprehension of emotional un- derstanding, one subbranch of the ability model of emotional intelligence (Mayer & Salovey, 1997). Currently, vignette items of hypothetical situ- 1 Chapter 1. General Introduction 2 ations are used for its measurement. In the conducted studies, emotional understanding was assessed in a natural setting and hypothetical situations were replaced with real ones. Participants were asked to predict their emo- tions and, later on, the same participants reported on their actual experi- ences. Predictions and experiences are now combined to calculate accuracy scores in terms of a low bias (i.e. Euclidean Distance) and high correspon- dence (i.e. profile correlation). This accuracy is related to abilities, traits, emotionally relevant criteria. The ability model aims to define emotional intelligence as intelligence in the content domain of emotions. To do so, performance should be evaluated as right or wrong, with right answers to be generally preferable. Since emo- tional understanding is defined as emotional knowledge, it is obvious that accurate knowledge should be worth thriving for. In this sense, Mayer and Salovey (1997) see emotional understanding as a prerequisite of emotional management, so that accurate knowledge can be used to effectively influ- ence emotions in ourselves and others. In other words, accuracy should be adaptive. Other lines of research have taken different approaches. The notion of depressive realism suggests that depression is associated with accurate ex- pectations which consolidate the disease (Alloy & Abramson, 1979). Others emphasized that a ’rosy view’ is preferable (Mitchell, Thompson, Peterson, & Cronk, 1997), and that there are certain biases in affective forecasting that can serve a purpose (Wilson & Gilbert, 2003). One of these biases is immune neglect, i.e. the lack of insight into mood repair processes and, therefore, the overestimation of the duration of emotions. This is supposed to be functional because it enforces to seek positive events and to avoid negative ones (Gilbert, Pinel, Wilson, Blumberg, & Wheatley, 1998). Generally, for the predictions of one’s own emotions, the adaptability of accuracy can be called into question. The current thesis explores this Chapter 1. General Introduction 3 domain with a special focus on predictions for different targets. Two studies were conducted. In the anxiety study, public officials predicted emotions in an important exam situation. The design is comparable to typical affective forecasting studies, and it focuses on a negative emotion toward a specific event. Results can be interpreted in context of affective forecasting studies, and it can be explored how interindividual differences in bias relate to cri- teria. Even if a general bias is functional, it probably should not be as high as possible. Also, an accurate view can be compared to a rosy one. In contrast, the work-group study focuses on the prediction of inter- personal feelings and relationship development. Following the EI rationale, emotional knowledge can be used to establish positive and effective relation- ships. This adds a social perspective to emotional predictions and allows to examineemotionalunderstandingastotheselfandothers. Thisisespecially important since the original conceptualization of ability EI emphasized this distinction. In both studies, accuracy of emotional predictions will be related to abilities,traits,andcriteria. Constructvalidityandincrementalvaluewillbe reported. Then, strengths and limitations will be discussed and possibilities for future research will be explored.