In Identity and Control, Harrison argued that social structure, culture and identity are continually evolving as individuals manage uncertainty by interacting. The resulting dynamics and changes in identity both control and can be controlled through the network. Generative network models, however, do not exhibit these properties. We argue that this is because they consider only the social and not the semantic network, and only use nodes at one level - people or organizations, not both. We ask whether Large Language Models (LLMs ) can be used to generate communities that behave as White described.

LLMs are increasingly utilized to generate synthetic data, providing valuable solutions in scenarios where real-world data is scarce, expensive, or privacy-sensitive. The hope is that LLMs will create artificial datasets that closely mimic real-world interactions and network structures so that it can be used for education, testing and training algorithms, and conducting behavioral research without compromising user privacy. The questions are: How realistic are these synthetic data? Are the networks realistic? How can such realism be assessed?

The AESOP-MOMUS system for generating and validating synthetic social media data is described. The resulting synthetic data are assessed for realism, and for exhibiting the properties discussed by White. We find default methods for using LLMs to create synthetic data generate messages that look real to humans, but that have very unrealistic semantic and social network properties. Harnessing the LLM to a network structure results in statis networks where the message content and agent identity don’t make sense given the agents network position. However, we find that by using principles defined in Identity and Control, more realistic data can be generated.

These principles include: 1) Structural equivalence – this is used to enable mutual changes across social and semantic networks. 2) Duality of structure – the individual’s role has causal priority over the messages they send. 3) Focus on uncertainty – uncertainty is created by incorporating inauthentic actors, low credibility information and news sites, and evolving events. 4) Narrative focus – meaning emerges from connections among actors, concepts, events. 5) Relational identity – actor identities are formed and maintained in relationships. 6) Control and agency – actors have agency which they employ when sending or refusing to send messages; however, the content is influenced by the emerging social and semantic networks. 7) Narrative construction – identities are constructed through narratives. In this process, many of these principles are extended from the original person or group social networks describe by White to both social and semantic networks. Finally, narrative is used as the basis for generating networks and identities and for interpreting them. How these play out in AESOP is illustrated showing generated networks and messages.

While more realistic, this synthetic data is limited as the dynamics are event based. Hence, there is only one context at any time. Future research should explore whether ambiguity handling in LLMs generated actor messages will support context flipping. Methods to improve the dynamics of identity and network are discussed.

This study examines how art galleries as cultural intermediaries structured exhibitions and shaped understanding of modern art during 1904-1915, a period of unprecedented artistic innovation preceding the institutionalization of modernist categories. Drawing on the Database of Modern Exhibitions (DoME) and associated visual archives, we investigate how galleries accomplished coherent exhibition curation despite lacking established category systems.

Our study reveals how network structures and cultural practices co-evolved during this pivotal period. We demonstrate that galleries developed classification practices through their interconnected exhibition decisions, creating informal networks of artistic association before formal movement categories emerged. The research combines two complementary data sources: historical exhibition records from DoME and computational analysis of exhibited artworks' visual features. This allows us to examine both the social networks formed through curatorial decisions and their relationship to emerging artistic styles.

Our findings illuminate how cultural intermediaries navigate evaluation in periods of categorical instability. We examine how galleries' classification choices were influenced by both network structures (relationships between artists and exhibitions) and cultural dynamics (evolving artistic conventions). This research aims to advance our understanding of classification processes in markets experiencing radical innovation, exploring how early social networks and cultural practices may shape the foundations of emerging category systems.

In Markets from Networks, Harrison White develops the concept of decoupling, notably in a chapter entitled “Embed and Decouple” (chapter 10). He also uses this concept in Identity and Control. This concept is a generalization of his theory of markets first presented in 1981 in an article entitled “Where do Markets come From? In this theory, firms in a situation of strong structural equivalence relative to suppliers or customers adjust to each other on the basis of price and quantity signals. These adjustments can lead to the stabilization of firms' positions in niches corresponding to product characteristics in terms of quality. In this case, reciprocal adjustments lead to relative consensus on quality assessment. The decoupling of a market is a process of emergence of an aggregate-level social formation from a network of dyadic relations between economic players. This social formation is structured by a type of order (a “discipline” in his language), which White calls an “interface”. Decoupling is a reciprocal of embeddedness, which is the dependence of a social formation on networks of dyadic relations. If generalized, this concept makes it possible to conceive (and empirically analyze) the emergence of a social entity in relation to others, which constitute or encompass it. In this paper, I will develop this possibility of generalization and take an example of the decoupling mechanisms of organizations in relation to persons and their networks. This example concerns research projects involving collaborations between academic laboratories and enterprises in the field of engineering. Many of these projects are decided and developed on the basis of chains of interpersonal relations, but the involvement of organizations (laboratory and enterprise) is reflected in the implementation of procedures for materialization (creation of models or documented digital models), formalization (contracts, agreements) or collectivization (information sharing, rotation of representatives in meetings) that reduce the effects of embeddedness in networks. Conversely, other processes, such as personalization (the monopolization of collaboration by one person in the enterprise or laboratory), weaken decoupling.

The existence of social fields – distinct sites of social competition and interaction around some common understanding of norms and goals – has long been recognized by sociologists. The mechanics underlying the emergence and development of these fields are, however, far less well understood. This is partly due to the way that these mechanics are shrouded by popular spatial models of fields which are ill-suited to processual questions about field evolution. Using data on 50 years of bill cosponsorship between United States senators, I construct a structural topic model which relates the social network interactions of these individuals through bills to the topical fields that these interactions emergently create. I show that the “speciation” of new topics can come from both intra-field and inter-field mechanisms which are amenable to discovery through the pairing of semantic and social network analysis. I describe how shifts in the mediating technologies of social network interaction – in this case the decline of the congressional committee system – influence the rate and kind of field evolution.

The increasing digitalization of our societies and the rapid development of artificial intelligence (AI) is tied to the emergence of professions related to what has been titled "data work". In the global South, hundreds of thousands of often precarious and poorly paid "data annotators" digitize, sort, and classify vast amounts of data.

In this presentation, I focus on the structure of the global "data work" market through a field study conducted between France and Madagascar in 2021-2022.Through a mixed methods, multi-site research design -- 300 questionnaires, and more than 200 interviews with various stakeholders: workers, managers, CEO, data-scientists and public servants -- I reconstructed an extensive network, mapping the exchange of data work related to the island.

Through an analysis that incorporates Harrison White's theory of markets and a stochastic blockmodel approach, I show that the exchange space is particularly structured by the chronic uncertainty that prevails there and the very low visibility that actors intervening within the network have with respect to the rest of the exchange space. In particular, I show that in order to minimize the risks associated with this uncertainty, a mechanism of double observation is set up within the space. Producers observe the signals sent by other producers in order to position their offer, but they also observe the signals sent by customers in order to select the projects they will accept. This analysis allows me to draw up a relational typology of the actors and to define the relational constraints that structure this space.

Harrison White was a towering figure in the development of network research. His passing gives us an opportunity to reflect on recent developments in the field: What might he say about them?

(1) White opposed rational choice modeling, and would probably be critical of Hedström’s program of individual-based analytical sociology. He saw individual identities not as driving forces behind social networks, but as constructed in network processes.

(2) He showed little interest in the study of network mechanisms like homophily, transitivity, and preferential attachment. His focus was on the identification of systematic network patterns specific to institutional fields.

(3) White seems to have ignored ego-centric networks in survey research. He was chiefly concerned with studying the meso-level of full network configurations.

(4) He would probably have been sympathetic to computational social science, with direct connections to pioneers of the field (Carley, DiMaggio, Mohr).

(5) As a trained physicist, White was sympathetic to developments in interdisciplinary network science. However, he was critical of the abstract mathematical modeling of social phenomena without substantially engaging with them.

(6) White would obviously have a strong affinity to relational sociology, but with little interest in theoretical forbears and ontological discussions.

(7) As a quantitative scholar, White was not interested in qualitative methods to study networks. However, the qualitative study of interaction / communication (e.g., with conversation analysis) and of meaning in networks fit his research aims more than qualitative interviews.

Among Harrison White's multiple contributions to economic sociology, the 1981 market model stands out as a cornerstone, emphasizing coordination through quality rather than price. In recent years, the rise of digital platform intermediaries has made economic activity more complex. The proposed presentation raises the question of coordination in the restaurants market since the arrival of these platforms. It extends and renews the interest of White’s model by showing that the role of platforms is more to support producers’ coordination through quality – thus aligning with his original perspective – than the matching of supply and demand, as standard economics would have it.

We analyze the effects of a popular booking and review platform on the dine-in restaurant market in the medium-sized city of Lille, France. In line with White’s framework, we conceptualize restaurants as a producers' market where multiple quality conventions — ordinary, advanced, and paradox — coexist with failure area. We instantiate these ideas through a combination of observations, interviews, web-scraping, and business data about 283 restaurants, following a sequential mixed-methods design. In particular, we use MRQAP analysis to show how dyadic similarities and differences in platform use, both by restaurants and by clients, predict their relative position in White’s market structure. We conclude that platforms rationalize firms’ practice of observing one another as a basis to make volume and quality decisions. The rise of digital platforms equips producers with devices that amplify their view of competitors, standardize their offerings, and support the stability of their business choices over time.