Aggregates of many judgments tend to outperform each of the individual judgments that compose the aggregate, termed the Wisdom of Crowds effect. Metacognition has played an understudied role in the efficacy of these crowds and so in a series of experiments I explore how metacognition and self-direction can be used to improve crowd wisdom. I first demonstrate empirically that individuals can leverage their metacognitive abilities to improve the performance of crowds when they are allowed to opt-in to questions of their choosing. I develop a Bayesian framework wherein latent contextual knowledge describes how crowd members make opt-in decisions to elucidate the relationship between these cognitive and metacognitive processes. I then show that metacognitive ability can be estimated by asking questions with no correct response options and create metacognitively wise crowds which achieve more accurate responses despite incorporating fewer crowd members. I discuss my contributions to a geopolitical forecasting competition in which I developed models that combine human and algorithmic judgments to create highly accurate forecasts of the future. In this competition, I evaluated the effects of attenuating forecaster self-direction in an applied setting. These findings collectively demonstrate the importance of metacognition in forming accurate aggregate judgments and clarify the underlying metacognitive processes involved in self-direction.

Introduction: The state of emergency department (ED) crowding in Pennsylvania has not previously been reported.

Methods: We assessed perceptions of ED crowding by surveying medical directors/chairs from Pennsylvania EDs in the spring of 2008.

Results: A total of 106 completed the questionnaire (68% response rate). A total of 83% (86/104) agreed that ED crowding was a problem; 26% (27/105) reported that at least half of admitted patients boarded for more than 4 hours. Ninety-eight percent (102/104) agreed that patient satisfaction suffers during crowding and 79% (84/106) stated that quality suffers. Sixty-five percent (68/105) reported that crowding had worsened during the past 2 years. Several hospital interventions were used to alleviate crowding: expediting discharges, 81% (86/106); prioritizing ED patients for inpatient beds, 79% (84/ 106); and ambulance diversion, 55% (57/105). Almost all respondents who had improved ED operations reported that it had reduced crowding.

Conclusion: ED crowding is a common problem in Pennsylvania and is worsening in the majority of hospitals, despite the implementation of a variety of interventions. [West J EmergMed. 2013;14(1):1–10.]

This dissertation addresses how contextualized expertise and task design can improve wisdom of the crowd estimates. The first two chapters apply the wisdom of the crowd to two related tasks that require spatial knowledge. The third chapter applies the wisdom of the crowd to a subset ranking task.

In Chapter 1, I investigate how framing effects impact the wisdom of the crowd. Participants selected tiles that either represented US states or African countries in two frames, present and absent. I constructed three wisdom of the crowd estimates: an unweighted average, a confidence-weighted average, and a wisdom of the crowd within estimate that combines an individual's responses across frames. I found that combining the estimates from the two frames resulted in an improved wisdom of the crowd estimate.

In Chapter 2, I build on the wisdom of the crowd application for a task that again requires spatial knowledge. Participants supplied a point estimate and a radius centered at that point estimate for where various US cities were located. Unweighted and radius-weighted wisdom of the crowd estimates were more accurate than most individuals, but the cognitive model-based wisdom of the crowd estimates tended to be even more accurate. I describe how using cognitive modeling that contextualizes expertise led to improved wisdom of the crowd estimates.

In Chapter 3, I present a new extension for the Thurstone model to partial ranking data. Ranking tasks have usually had participants rank all items, but I present two different types of partial ranking tasks where either an experimenter or a participant selects the items to be ranked. I demonstrate how the Thurstone model can be used to generate wisdom of the crowd estimates, and speculate how other partial ranking tasks can be developed to better elicit diverse estimates from the crowd.

In all, these chapters detail specific applications of the wisdom of the crowd effect that better contextualize expertise, elicit multiple meaningful estimates from the same individual, and improve diversity. These methods are used in conjunction with cognitive modeling to produce improved wisdom of the crowd estimates.

Despite decades of research, significant mysteries remain concerning two

fundamental aspects of human visual processing: how we perceive moderately complex line shapes such as letters of the alphabet, and how peripheral vision differs from foveal vision. This dissertation comprises studies that address both these topics, using experiments with letter recognition and visual "crowding"--the deleterious influence of flanking objects on identification of a target.

Letters are well-studied psychophysical stimuli that strike a balance between degenerate patterns such as Gabor patches or gratings, with more uncontrolled stimuli such as natural images. Letters contain distinct spatial features that are more straightforward to quantify than the characteristics of natural scenes, yet have a sufficient richness of information to probe the general-purpose mechanisms of object recognition.

The last decade has seen a renewed interest in perception in the peripheral visual field, with a particular emphasis on the phenomenon of crowding. Many researchers acknowledge that this curious effect provides a unique window into the process of object recognition. Since there is such a marked difference between the periphery, where crowding is potent, and the fovea, where it is nearly absent, crowding sheds light on the dichotomy between visual processing in the high resolution fovea versus the information-limited periphery. Understanding crowding may illuminate general principles of biological visual processing.

Crowdsourcing is a powerful approach to solve some of the biggest problems today - social and scientific. To harness the potential of crowdsourcing, researchers have often relied and successfully used crowd marketplaces such as Amazon Mechanical Turk. However, in spite of a large worker base and past successes, these platforms come with their own limitations. 40% of all the tasks are spam on Mechanical Turk; and it’s also difficult to accomplish tasks which require a broader participation base, or expert knowledge. Citizen crowd, at the same time, involves distributed group of people with diverse background and skills; who might organize to accomplish tasks or solve problems without engaging with paid crowd platforms. The motivation for them lies in reasons other than micro-payments.

In this dissertation, I explore problems of varying size and duration, where mobilizing the citizen crowd can achieve worthwhile goals. To investigate the possibilities with citizen crowdsourcing, I built systems and ran initiatives that require different level of expertise. This thesis highlights three projects, ranging from seconds of micro contributions to months of highly intellectual contributions. To begin with, we built a mobile lock screen application, that supports completion of quick microtasks in short bursts of time - each time someone unlocks their phone. Next, we launched first of its kind large scale distributed and time-critical campaign in India; where we mobilized the crowd and studied aspects around team building, collaboration, incentives and communication channel; to achieve a common goal. Finally, we ran a first of its kind "research-at-scale" initiative, that explores the possibility of crowdsourcing large open ended problems - by connecting professors with students from around the world - providing access and educational value. All these initiatives have engaged thousands of people from around the world, and we can conclude that it is possible to mobilize the citizen crowd to accomplish worthy goals.

Abstract

The Birth of the Mob: Representations of Crowds in Archaic and Classical Greek Literature

by

Justin Jon Schwab

Doctor of Philosophy in Classics

University of California, Berkeley

Professor Leslie Kurke, Chair

This dissertation surveys the representation of crowds and related phenomena in Homer, the Attic tragedians, and Aristophanes.

The first chapter begins by noting that while recent scholarship has explored the role of the crowd in ancient Roman history and literature, virtually no similar work has been done in archaic and classical Greek studies. Admittedly, Greek poleis were on a much smaller scale than was Rome, and it may be for this reason that classical scholars have assumed "the" crowd is not a feature of ancient Greek society. In order to explain why this absence of study is due to a limited understanding of what crowds are, I survey the development of crowd theory and mass psychology in the modern era. I adopt the model of Elias Canetti's Crowds and Power, which studies crowds as part of a spectrum of group behavior, ranging from small "packs" to imagined crowds at the level of a nation. Under this expanded model, I argue that crowds are universal human phenomena whose representations in archaic and classical Greek literature are fruitful objects of study. The chapter ends with a brief survey of "crowd words" to be examined, including homilos, ochlos, homados and thorubos.

The second chapter studies crowds in Homer through a close reading of several words and passages. The two crucial words for this study are homilos and homados, which refer respectively to a crowd and the distinctive noise it makes. I survey the homilos in the Iliad as a background of anonymous figures against which elite figures display their excellence, before arguing that the suitors in the Odyssey are the closest Homer comes to representing a crowd. Individually elite, they nonetheless are reduced to the status of a mob by the fact of their aggregation.

The third chapter examines the crowd in tragedy. I argue that the crowd looms as an offstage threat to the elite characters depicted onstage, most obviously in such plays as Sophocles's Ajax and Euripides's Andromache and Orestes, but to some extent in almost every surviving tragedy. In this chapter, the word ochlos (not yet present in Homer) is the key crowd-term, although homilos and other words are also present. The works of Euripides are particularly rife with descriptions of crowds, and my survey illuminates just how central the topic was to his work, in a reflection of the troubled politics of his era.

The fourth chapter examines the discourse on the crowd in Aristophanes. I demonstrate that the comedian's work is highly concerned with crowds and other groupings of people. Athens during the Peloponnesian war was crowded, not only due to the siege but in mentality and dramatic representation. To many of Aristophanes's characters, the improper aggregation of bodies is just one symptom of the general disintegration of society and decline of traditional morality. Where in tragedy the crowd must remain offstage, comedy can also bring crowds onto the stage, in such scenes as the opening of the Acharnians.

I close with a Postscript presenting two quotes of Plato, from the Republic and the Laws, whose descriptions of crowd behavior and its effect on individuals take on new significance in light of the deep history of the representation of crowds which this dissertation explores.

Objective: Crowding in the emergency department (ED) impacts a number of important quality and safety metrics. We studied ED crowding measures associated with adverse events (AE) resulting from central venous catheters (CVC) inserted in the ED, as well as the relationship between crowding and the frequency of CVC insertions in an ED cohort admitted to the intensive care unit (ICU).

Methods: We conducted a retrospective observational study from 2008–2010 in an academic tertiary care center. Participants undergoing CVC in the ED or admitted to an ICU were categorized by quartile based on the following: National Emergency Department Overcrowding Scale (NEDOCS); waiting room patients (WR); ED patients awaiting inpatient beds (boarders); and ED occupancy (EDO). Main outcomes were the occurrence of an AE during CVC insertion in the ED, and deferred procedures assessed by frequency of CVC insertions in ED patients admitted to the ICU.

Results: Of 2,284 ED patients who had a CVC inserted, 293 (13%) suffered an AE. There was no association between AEs from ED CVCs and crowding scales when comparing the highest crowding level or quartile to all other quartiles: NEDOCS (dangerous crowding [13.1%] vs other levels [13.0%], P = 0.98); number of WR patients (14.0% vs 12.7%, P = 0.81); EDO (13.0% vs 12.9%, P = 0.99); and number of boarding patients (12.0% vs 13.3%), P = 0.21). In a cohort of ED patients admitted to the ICU, there was no association between CVC placement rates in the ED and crowding scales comparing the highest vs all other quartiles: NEDOCS (dangerous crowding 16% vs all others 16%, P = 0.97); WR patients (16% vs 16%, P = 0.82), EDO (15% vs. 17%, P = 0.15); and number of boarding patients (17% vs 16%, P = 0.08).

Conclusion: In a large, academic tertiary-care center, frequency of CVC insertion in the ED and related AEs were not associated with measures of crowding. These findings add to the evidence that the negative effects of crowding, which impact all ED patients and measures of ED performance, are less likely to impair the delivery of prioritized time-critical interventions.