In the era of hybrid and digital work, the increasing adoption of digital communication tools has led to the accumulation of vast amounts of unstructured data, providing organizations with new opportunities to understand the intricacies behind the wants, needs, and behaviors of employees. This can be crucial to improving people's well-being and overall organizational success. This thesis investigates how digital communication data, such as asynchronous (i.e., email) and synchronous (i.e., audio/video calls), can be leveraged to study organizational behavior and social capital within firms. The work, as a first deliverable, provides the design and implementation of a comprehensive and robust framework which consists in the acquisition, processing, and analysis of large-scale digital communication, organizational and survey based data. The framework integrates multiples steps: from data extraction and anonymization to network modeling and validation. As a result, this study introduces a novel large-scale dataset related to two real-world banking companies, including 3,108 and 1,149 employees for Company A and Company B, respectively. While preserving data privacy, this work integrates socio-demographic and organizational attributes, employee surveys with large-scale digital interaction organizational networks, including email messages, and audio/video calls. Building on this novel and unique dataset, in addition to the methodological contribution provided by the framework, this work presents three empirical contributions as research case studies. The first focuses on uncovering hidden informal communication networks and comparing them to formal organizational structures by applying state-of-the-art community detection algorithms. Our findings show that, while informal communities may overlap with the formal structure to some extent, they often diverge in ways that may reveal organizational dynamics, communication flow, and interdepartmental collaboration that are not apparent from the formal structure alone. The second research case study investigates the effects of social capital, analyzed through digital communication, on employee well-being and work performance. The results indicate that, while increased email connectivity is positively associated with employee performance and well-being, the benefits diminish beyond a certain level of centrality. Employees with low centrality experience notable gains in engagement, satisfaction, and performance as their connectivity increases. However, highly central employees do not continue to see performance gain, suggesting the presence of an optimal range of digital communication activity. To our knowledge, this is the first study simultaneously exploring the performance and employee well-being in relation to digital communication patterns, offering a comprehensive vision of the integration of these two perspectives. Finally, the third case study examines mixed-modal communication (e.g., email and audio/video calls), showing that synchronous and asynchronous channels complement each other differently across socio-demographic variables, indicating that employees have different communication behaviors within the company. Furthermore, by applying supervised machine learning models, the study achieves high accuracy in distinguishing managers based on communication features. Feature importance analysis through SHAP values further uncovers behavioral signatures of managerial activity. Together, the comprehensive framework and these studies, aim to pave the way for future research on exploring how computational social science methods, such as Organizational Network Analysis based on digital communication data, can foster research on organizational behavior and enhance human resource management practices. This work contributes both methodological and practical implications for human resource management and organizational behavioral studies.
Nell'era del lavoro ibrido e remoto, la crescente adozione di strumenti di comunicazione digitale ha contribuito all'accumulo di grandi quantità di dati non strutturati, offrendo alle aziende nuove opportunità per comprendere a fondo richieste, necessità e comportamenti dei dipendenti. Questa opportunità può essere utilizzata per migliorare il benessere delle persone e il successo dell'organizzazione. La tesi in questione esplora come i dati di comunicazione digitale, sia asincrona con email, che sincrona con chiamate audio/video, possano essere sfruttati per studiare il comportamento organizzativo e il capitale sociale all'interno delle imprese. Come primo risultato, questo lavoro propone la progettazione e implementazione di un framework completo e robusto che consiste in acquisizione, elaborazione e analisi di dati relativi alla comunicazione digitale, organizzativi e basati su survey in larga scala. Il framework è costituito da diverse fasi, che vanno dall'estrazione e anonimizzazione dei dati alla modellazione e validazione delle reti organizzative e di comunicazione digitale. Come risultato, questo studio introduce un nuovo large-scale dataset relativo a due aziende bancarie in ambito real-world, comprendente rispettivamente di 3,108 e 1,149 dipendenti per Azienda A e Azienda B. Preservando la privacy dei dipendenti, questa ricerca integra dati socio-demografici e organizzativi, survey compilate da dipendenti e large-scale digital social networks usando i log di messaggi relativi a email e chiamate audio/video. In aggiunta all'introduzione della metodologia sviluppata e il dataset acquisito, la tesi presenta tre contributi empirici sotto forma di casi di studio. Il primo analizza le reti informali di comunicazione digitale nascosti e le confronta con le strutture organizzative formali, applicando algoritmi di community detection non supervisionati. I risultati mostrano che, sebbene esistano sovrapposizioni tra comunità informali e struttura formale, spesso emergono differenze che rivelano dinamiche organizzative, flussi di comunicazione e collaborazioni interdipartimentali non evidenti dalla sola struttura ufficiale dell'azienda. Il secondo caso di studio esamina gli effetti e impatti del capitale sociale, analizzato tramite i dati di comunicazione digitale, sul benessere e sulle performance dei dipendenti. I risultati indicano che un aumento della connettività nelle email è associato positivamente alla performance e al benessere dei lavoratori, ma i benefici tendono a non essere più evidenti oltre un certo livello di centralità. I dipendenti con bassa centralità traggono maggiori benefici in termini di coinvolgimento, soddisfazione e prestazioni, mentre quelli con alta centralità non mostrano ulteriori miglioramenti, suggerendo l'esistenza di un livello ottimale di attività comunicativa digitale. Sulla base della nostra conoscenza, questo è il primo studio che esplora simultaneamente la performance e il benessere dei dipendenti in relazione ai pattern di comunicazione digitale, offrendo una visione integrata di entrambe le prospettive. Il terzo caso di studio analizza la comunicazione multimodale (es. email e chiamate audio/video), mostrando il comportamento dei dipendenti attraverso canali sincroni e asincroni e come questi si integrano in modo differente in base a variabili socio-demografiche e organizzative. Inoltre, attraverso l'applicazione di modelli di machine learning supervisionati, lo studio mostra un'elevata accuratezza nell'identificazione dei manager sulla base delle features estratte dai social networks digitali. A tal proposito, attraverso SHAP, si è valutata anche l'importanza delle features portando all'individuazione di pattern comportamentali tipici dell'attività manageriale attraverso i canali digitali. Nel complesso, il framework proposto e i tre studi contribuiscono allo stato dell'arte, mostrando come metodi di computational social science, come l'Organizational Network Analysis basata su canali di comunicazione digitale, possano supportare lo studio del comportamento organizzativo e migliorare le pratiche di gestione delle Risorse Umane. Questa tesi apporta implicazioni sia metodologiche che pratiche per la gestione delle risorse umane e per gli studi del comportamento aziendale.
Beneath the surface: unveiling real-world organizations hidden digital communication networks
Di PERNA, LEONARDO
2024/2025
Abstract
In the era of hybrid and digital work, the increasing adoption of digital communication tools has led to the accumulation of vast amounts of unstructured data, providing organizations with new opportunities to understand the intricacies behind the wants, needs, and behaviors of employees. This can be crucial to improving people's well-being and overall organizational success. This thesis investigates how digital communication data, such as asynchronous (i.e., email) and synchronous (i.e., audio/video calls), can be leveraged to study organizational behavior and social capital within firms. The work, as a first deliverable, provides the design and implementation of a comprehensive and robust framework which consists in the acquisition, processing, and analysis of large-scale digital communication, organizational and survey based data. The framework integrates multiples steps: from data extraction and anonymization to network modeling and validation. As a result, this study introduces a novel large-scale dataset related to two real-world banking companies, including 3,108 and 1,149 employees for Company A and Company B, respectively. While preserving data privacy, this work integrates socio-demographic and organizational attributes, employee surveys with large-scale digital interaction organizational networks, including email messages, and audio/video calls. Building on this novel and unique dataset, in addition to the methodological contribution provided by the framework, this work presents three empirical contributions as research case studies. The first focuses on uncovering hidden informal communication networks and comparing them to formal organizational structures by applying state-of-the-art community detection algorithms. Our findings show that, while informal communities may overlap with the formal structure to some extent, they often diverge in ways that may reveal organizational dynamics, communication flow, and interdepartmental collaboration that are not apparent from the formal structure alone. The second research case study investigates the effects of social capital, analyzed through digital communication, on employee well-being and work performance. The results indicate that, while increased email connectivity is positively associated with employee performance and well-being, the benefits diminish beyond a certain level of centrality. Employees with low centrality experience notable gains in engagement, satisfaction, and performance as their connectivity increases. However, highly central employees do not continue to see performance gain, suggesting the presence of an optimal range of digital communication activity. To our knowledge, this is the first study simultaneously exploring the performance and employee well-being in relation to digital communication patterns, offering a comprehensive vision of the integration of these two perspectives. Finally, the third case study examines mixed-modal communication (e.g., email and audio/video calls), showing that synchronous and asynchronous channels complement each other differently across socio-demographic variables, indicating that employees have different communication behaviors within the company. Furthermore, by applying supervised machine learning models, the study achieves high accuracy in distinguishing managers based on communication features. Feature importance analysis through SHAP values further uncovers behavioral signatures of managerial activity. Together, the comprehensive framework and these studies, aim to pave the way for future research on exploring how computational social science methods, such as Organizational Network Analysis based on digital communication data, can foster research on organizational behavior and enhance human resource management practices. This work contributes both methodological and practical implications for human resource management and organizational behavioral studies.| File | Dimensione | Formato | |
|---|---|---|---|
|
PhD_Thesis_LeonardoDiPerna.pdf
accessibile in internet per tutti
Dimensione
29.68 MB
Formato
Adobe PDF
|
29.68 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/244957