2018-2019 Spring Colloquia

ABSTRACT: How to understand diverse forms of social action ranging from high-risk street protest to fandom on livestreaming platforms? Neither the classical logic of collective action nor recent formulations of connective action provide adequate answers. I will argue that a theory of the logic of performative action can capture the meanings of social action in ways that elude the logic of collective or connective action. To explicate the logic of performative action, I start by tracing it back to Weber’s social theory. I will then show how the same logic can explain both death-defying political radicalism and entertainment behavior on Chinese and American social media.

ABSTRACT: Replication and transparency are increasingly important in bolstering the credibility of political science research, yet open science tools are typically designed for experiments. For observational studies, current replication practice suffers from an important pathology: just as researchers can often “p-hack’’ their way to initial findings, it is often possible to “null hack’’ findings away through specification and case search. We propose an observational open science framework that consists of extending the original time series, independent data collection, pre-registration, multiple simultaneous replications, and collaborators with mixed incentives. We apply the approach to three studies on “irrelevant’’ events and voting behavior. Each study replicates well in some areas and poorly in others. Had we sought to debunk any of the three with ex post specification search, we could have done so. However, our approach required us to see the full, complicated picture. We conclude with suggestions for future refinements to our approach.

ABSTRACT: Attrition and conditioning can introduce special biases unique to panel surveys, compromising their usefulness as a measurement tool. Using validated turnout data matched to the Pew American Trends Panel, I show that panel surveys are as accurate as simple cross-sectional surveys in measuring voter turnout. Using validated turnout data on 5,503 Panelists, I find that panel attrition among non-voters is small and ignorable once weights are applied. To identify the effect of panel conditioning, I compared a control group of panelists recruited after the 2014 general election and a treatment group of panelists that had been in the panel for nearly 2 years before the 2014 election. Using prognosis score matching, the two groups were matched on basic demographics and turnout variables from the voter file. Participating in the panel made panelists about 2 percentage points more likely to vote in the 2014 election, but this higher level of participation is counteracted by decreased over-reporting of voting among panelists. Once all these panel effects are taken into account, estimates of 2014 voting rates from long-term ATP Panelists are about the same as they would be from a simple crosssectional survey.

Chris was a member of PPRG when he was here at Stanford in graduate school for psychology, and he is back this year as a fellow at the Center for Advanced Study in the Behavioral Sciences (CASBS). He will be coming to talk about his research on social interventions to change behavior on large scales in real-world settings

ABSTRACT: This talk uses Dynamic Human Centered Communication Systems Theory, Gibson’s ecological perception theory, and Dynamic Systems Theory to reconceptualize and analyze the emergence of presence and behavioral change at two time scales (development and learning) in virtual game worlds. First, an analysis is reported that attempts to determine the state space of presence when pursuing a goal in a virtual world and the stability of the different kinds of presence within the system and within individuals. Next, a qualitative analysis of 34 participants playing Grand Theft Auto IV (GTA) for 30 minutes was undertaken to identity the state space of the GTA-humanenvironment system, the system and individual stabilities of the GTA behavioral attractors, and the process of destabilizing real world attractors and stabilizing new GTA attractors over developmental and learning time scales. In addition, the influence of two biological initial conditions on the trajectory of attractor stability is examined.

ABSTRACT: The talk will touch on two topics. The first is recent progress on automatically designing adaptive experiments. this is a really old idea, but new progress in machine learning and statistics is starting to make it more feasible for non-experts to use tools for optimal experiment design. given work on survey design, we can think through the most interesting ways to use these new tools. Then, time permitting, I will talk about models of language use. I’ll discuss work on the rational speech acts framework for probabilistic modeling of natural language pragmatics, with examples such as hyperbole and vagueness, and then attempts to learn semantics for natural language via reference games.

ABSTRACT: Financial markets have been disrupted and reconfigured by major technological advances three times in the last 150 years. First by the stock ticker, whose popularity spread nearly instantly once it had been invented in 1886. Then, in the 1980s, a hundred years later, by the computer, with screens replacing ticker tape and enabling electronic trading. The last major disruptive shift is the move to trading by algorithms. It has been in full force only since about 2005 in various markets, and yet created plenty of upheaval in trading professions and among regulators of financial trading. What sort thing is an algorithm that trades? How did we get there—what genealogy of preferences leads to the present coming to be what it is now, semi-autonomous markets? And what happens to the flat interaction-level market when algorithms are the counterparties of trades? The paper highlights aspects of a transition from microinstitutional to mediainstitutional financial markets and how this manifests itself on a system-, organizational-, agentic and legal level.