Social-influence processes not only affect the rate at which behaviors spread but can also decouple adoption behavior from individual preferences, and thereby bring about unexpected collective outcomes that cannot be predicted on the basis of the initial likes and dislikes of the individuals involved. However, the conditions under which social influence can lead to such decoupling are not well understood. We identify a social-influence mechanism that widens individuals’ behavioral repertoires and breaks the link between individuals’ initial preferences and the collective outcomes they jointly bring about. We test the micro-level assumptions of the mechanism in the context of cultural choices on Spotify, combining topic modeling with traditional statistical matching to cultural change. agent-based simulation estimate peer-to-peer influence effects from digital trace data. We then use agent-based simulations to examine the macro-level consequences of “wide” social influence and its importance for explaining cultural change.

The digital and computational revolutions have improved the prospects for analyzing the dynamics of large groups of interacting individuals. Digital trace data provide the type of large-scale, time-stamped, and granular information on social interactions that is needed to feasibly conduct research on social influence in non-experimental settings and to distinguish social influence effects from the confounding effects of homophily. This chapter reviews three concrete ways in which machine learning can improve the estimation of social influence effects from observational digital trace data. These computational approaches (a) make high-dimensional information about individuals accessible for analysis, (b) infer latent confounders from the structure of large-scale social networks, and (c) facilitate large-scale annotation of measures that can serve as instruments for causal identification

One explanation for why people accept ideologically welcome misinformation is that they are insincere. Consistent with the insincerity hypothesis, past experiments have demonstrated that bias in the veracity assessment of publicly reported statistics and debunked news headlines often diminishes considerably when accuracy is incentivized. Many statements encountered online, however, constitute previously unseen claims that are difficult to evaluate the veracity of. We hypothesize that when confronted with unfamiliar content, unsure partisans will form sincere beliefs that are ideologically aligned. Across three experimental studies, 1,344 conservative and liberal US participants assessed the veracity of 20 politically sensitive statements that either confirmed or contradicted social science evidence only known to experts. As hypothesized, analyses show that incentives failed to correct most ideological differences in the perceived veracity of statements. 66 to 78% of partisan differences in accuracy assessment persisted even when monetary stakes were raised beyond levels in prior studies. Participants displayed a surprising degree of confidence in their erroneous beliefs, as bias was not reduced when participants could safely avoid rating statements they were unsure about, without monetary loss. These findings suggest limits to the ability of disciplining interventions to reduce the expression of false statements, because many of the targeted individuals sincerely believe them to be true.

How does the ideological segregation of online networks impact the spread of misinformation? Paststudies have found that homophily generally increases diffusion, suggesting that partisan news,whether true or false, will spread farther in ideologically segregated networks. We argue that networksegregation disproportionately aids messages that are otherwise too implausible to diffuse, thusfavoring false over true news. To test this argument, we seeded true and false informational messagesin experimental networks in which subjects were either ideologically integrated or segregated,yielding 512 controlled propagation histories in 16 independent information systems. Experimentalresults reveal that the fraction of false information circulating was systematically greater inideologically segregated networks. Agent-based models show robustness of this finding acrossdifferent network topologies and sizes. We conclude that partisan sorting undermines the veracity ofinformation circulating on the Internet by increasing exposure to content that would otherwise notmanage to diffuse.

Are the best-paying jobs with the highest prestige done by individuals of great intelligence? Past studies find job success to increase with cognitive ability, but do not examine how, conversely, ability varies with job success. Stratification theories suggest that social background and cumulative advantage dominate cognitive ability as determinants of high occupational success. This leads us to hypothesize that among the relatively successful, average ability is concave in income and prestige. We draw on Swedish register data containing measures of cognitive ability and labour-market success for 59,000 men who took a compulsory military conscription test. Strikingly, we find that the relationship between ability and wage is strong overall, yet above €60,000 per year ability plateaus at a modest level of +1 standard deviation. The top 1 per cent even score slightly worse on cognitive ability than those in the income strata right below them. We observe a similar but less pronounced plateauing of ability at high occupational prestige. 

Theories of urban scaling have demonstrated remarkable predictive accuracy at aggregate levels. However, they have overlooked the stark inequalities that exist within cities. Human networking and productivity exhibit heavy-tailed distributions, with some individuals contributing disproportionately to city totals. Here we use micro-level data from Europe and the United States on interconnectivity, productivity and innovation in cities. We find that the tails of within-city distributions and their growth by city size account for 36–80% of previously reported scaling effects, and 56–87% of the variance in scaling between indicators of varying economic complexity. Providing explanatory depth to these findings, we identify a mechanism—city size-dependent cumulative advantage—that constitutes an important channel through which differences in the size of tails emerge. Our findings demonstrate that urban scaling is in large part a story about inequality in cities, implying that the causal processes underlying the heavier tails in larger cities must be considered in explanations of urban scaling. This result also shows that agglomeration effects benefit urban elites the most, with the majority of city dwellers partially excluded from the socio-economic benefits of growing cities.

Analytical sociology is beginning to embrace a digital revolution in the collection and analysis of social data and is increasingly drawing on tools from computational social science (CSS) to pursue its goals of mechanism-based explanation of aggregate outcomes. In this chapter, we highlight the ways in which analytical sociologists are using CSS tools to further social research. Using agent-based modeling, large-scale online experiments, digital trace data, and natural language processing, analytical sociologists are identifying how large-scale properties of social systems emerge from the complex interactions of networked actors at lower scales. At the same time, we provide a perspective on how CSS techniques can be successfully deployed in social research, including ways in which they can be productively combined. Computational tools, when applied using a theory-grounded approach, offer sociologists a chance to transcend the limitations of the dominant survey-research paradigm and finally address “big” sociological questions about, for example, the nature of culture, the emergence of inequality, and the dynamics of segregation. We also discuss how computational social scientists can take cues from analytical sociology to further hone their own research and methods in the service of theoretically grounded, mechanism-based explanations, moving beyond theoretically thin descriptions or predictions of micro- and macro-level outcomes.

The “transportability” of laboratory findings to other instances than the original implementation entails the robustness of rates of observed behaviors and estimated treatment effects to changes in the specific research setting and in the sample under study. In four studies based on incentivized games of fairness, trust, and reciprocity, we evaluate (1) the sensitivity of laboratory results to locally recruited student-subject pools, (2) the comparability of behavioral data collected online and, under varying anonymity conditions, in the laboratory, (3) the generalizability of student-based results to the broader population, and (4), with a replication at Amazon Mechanical Turk, the stability of laboratory results across research contexts. For the class of laboratory designs using interactive games as measurement instruments of prosocial behavior we find that rates of behavior and the exact behavioral differences between decision situations do not transport beyond specific implementations. Most clearly, data obtained from standard participant pools differ significantly from those from the broader population. This undermines the use of empirically motivated laboratory studies to establish descriptive parameters of human behavior. Directions of the behavioral differences between games, in contrast, are remarkably robust to changes in samples and settings. Moreover, we find no evidence for either anonymity effects nor mode effects potentially biasing laboratory measurement. These results underscore the capacity of laboratory experiments to establish generalizable causal effects in theory-driven designs.

Globalization makes cross-national encounters increasingly common. Hesitant cooperationacross national, ethnic, and cultural boundaries, however, undercuts the microlevel stabilizers of global integration and, most importantly, the willingness to share with and place trust in members of other social groups. In a 109-country online experiment, we convey information on interaction partners’ nationalities to indicate membership in a broader in- or out-group, cultural distance, and perceived material neediness—or status differences more generally—to 1,674 participants in incentivized games of generosity (dictator game) and trust (trust game). We find consistent evidence for in-group favoritism and—against this benchmark—demonstrate that individuals across the globe share more with but place less trust in interaction partners from poor countries and that cultural distance moderates this status effect.

The advantages of groups over individuals in complex decision-making have long interested scientists across disciplinary divisions. Averaging over a collection of individual judgments proves a reliable strategy for aggregating information, particularly in diverse groups in which statistically independent beliefs fall on both sides of the truth and contradictory biases are cancelled out. Social influence, some have said, narrows variation in individual opinions and undermines this wisdom-of-crowds effect in continuous estimation tasks. Researchers, however, neglected to study social-influence effects on voting in discrete choice tasks. Using agent-based simulation, we show that under voting — the most widespread social decision rule — social influence contributes to information aggregation and thus strengthens collective judgment. Adding to our knowledge about complex systems comprised of adaptive agents, this finding has important ramifications for the design of collective decision-making in both public administration and private firms.