Skip to main content

The Implications of Selective Information Sampling for Individual and Collective Judgments

Periodic Reporting for period 2 - InfoSampCollectJgmt (The Implications of Selective Information Sampling for Individual and Collective Judgments)

Reporting period: 2019-11-01 to 2021-04-30

I set to understand the mechanisms leading to the polarization of attitudes across social groups. This is important because an excessive polarization of attitudes prevents people from understanding each other. This creates a challenge to the democratic debate and imposes huge costs on our societies.
A number of commentators have proposed that ‘filter bubbles’ are the key explanation for the polarization of political attitudes. People would live in information silos, being mostly exposed to ideas consistent with their opinions and lacking exposure to contrarian ideas. The ‘filter bubble’ conjecture is a sampling explanation: it focuses on how past experiences and the social environment affect the information people sample. Like other sampling explanations, including those I have developed in my past research, it does not specify the cognitive mechanisms that translate information into attitudes.
My ERC project incorporates insights from cognitive and social psychology into this sampling explanation. The idea is to study how mechanisms such as confirmation biases or motivated cognition affect the predictions of the filter bubble hypothesis and other sampling-based mechanisms.
What is groundbreaking about the project is that it brings together two classes of mechanisms (sampling-based and information processing-based) that have been treated in isolation. The project analyzes how the interaction between these mechanisms affects the dynamics of individual judgments and attitudes, collective judgments and attitudes, and finally, the distribution of attitudes over social networks.
My plan for the first 18 months was to work on three work packages:
Work package 1: I have developed models that combine sampling and information processing components to explain judgments about the variability of categories (with E. Konovalova) and social groups in particular and to explain evaluative judgments toward categories (with T. Woiczyk). I have analysed the consequences of adaptive sampling for the chooser’s wellbeing (with Y. Kareev, J. Avrahami and A. Gelastopoulos). I have developed a model that shows that when decision makers learn about the values of alternatives from their own experiences, they are subject to a sampling bias against alternatives that improve with practice (with J. Denrell). I have developed a model about how people aggregate review information expressed on different scales to inform their sampling decisions (with R. Hosseini). I have developed models how mental categories affect inferences and evaluations (with E. Konovalova, M. Hannan, B. Kovács, T. Woiczyck, F. Sump)
Workpackage 2: I have tested the predictions of the models developed with Konovalova, Woiczyk, Hosseini, Hannan, Kovacs, Sump in 15+ online experiments. I have also run experiments that examine how the sampling mode (active versus passive) affects the processing of information (with J. Gisbert). I have collected a large dataset of tweets published by Spanish elected officials to analyze the extent they engaged in adaptive sampling when selecting the topics of their tweets and how this can explain gender differences in issue attention (with N. Schöll and A. Gallego). I have trained myself in using deep learning models for natural language processing to use state-of-the-art techniques to analyze twitter data (the new ‘transformer’ class of models for natural language understanding, such as BERT).
Workpackage 3: I have developed two new models to explain attitude polarization (with E. Konovalova and N. Schöll) and begun to test some of their predictions in online experiments. I have also started to use the twitter dataset to validate some of the model assumptions. I have developed a new model that explores the implications of ranking algorithms for the popularity of news sources, and tested model predictions in online experiments (with F. Germano and V. Gómez)
In my application, I had identified three limitations of the state-of-the art
1. How information sampling biases and information processing biases interact in producing judgment biases is unclear.
2. The ecological validity of the premises of sampling-based theories is unclear.
3. The implications of the sampling-based theories for collective judgments and the distribution of beliefs and attitudes over social networks are unclear.
The projects with E. Konovalova, T. Woiczyk, J. Gisbert, R. Hosseini and F. Sump all contribute to point 1. Projects with Hosseini (The Scale Effect: How Rating Scales Affect Product Evaluation) and Woiczyk (Evaluating Categories from Experience: The Simple Averaging Heuristic) lead to better understanding how people integrate sampled information in forming attitudes and evaluative judgments and how this in turn affect subsequent sampling behaviour. The project with Woiczyk and Sump (Distinctiveness through Categories: Generalization and the Exploration of Novel Options) analyses how common or distinct labelling affect people’s propensity to sample exploratory alternatives and learn about their values. The project with Gisbert (How selective access to financial information affects how investors learn) experimentally investigates how the sampling mode affect the accuracy of information integration. The project with E. Konovalova (An Information Sampling Explanation for the In-Group) shows that the effect of the sampling asymmetry persists under a broad set of assumptions regarding the interaction between sampling and information processing.
The analysis of Twitter data (with N. Schöll and A. Gallego) contributes to point 2. It has demonstrated that the basic assumptions of sampling models hold in naturally occurring environments and that the sampling approach contributes to explaining important real world phenomena (systematic differences in the topics about which what politicians from different social groups write). In the next few months, most of my team’s effort will focus on further analyzing these data to measure the properties of the samples of information twitter users obtain as they interact via the platform in order to validate both the assumptions and predictions of sampling model
The projects with E. Konovalova, N. Schöll, V. Gómez and F. Germano contribute to point 3. The project with E. Konovalova and N. Schöll (Asymmetric Feedback Can Contribute to Polarization) proposes that the nature of the feedback people obtain when they post content on social media can contribute to polarization. The project with F. Germano and V. Gómez (The Few-Get-Richer Effect) has shown that how recommendation systems affect sampling behaviour has the potential to make incorrect beliefs and information popular and casts lights on how fake news can become widespread. In the next period, I am planning to put to test the predictions of the project with E. Konovalova and N. Schöll in collective behaviour experiments. I am also planning to develop the implications of the few-get-richer effect for collective well-being using computational models and design further tests of the predictions of the model.