Research

1. [Title hidden to ensure blind review -- a piece on Generative AI and Organizational Skills]

with: Dogukan Yilmaz, Vibha Gaba, & Henrich Greve

Status: R&R at Management Science

Paper in a nutshell

This study examines how Generative AI reshapes organizational functions. Using a quasi-experimental design around ChatGPT's introduction, it reveals selective skill deprioritization, notably a decline in demand for Task Division and Information Provision.

2. [Title hidden to ensure blind review -- a piece on Generative AI and Hierarchies]

with: Dogukan Yilmaz, Vibha Gaba, & Henrich Greve

Status: Under Review at Organization Science

Paper in a nutshell

This paper examines how Generative AI reshapes organizational hierarchies by reducing the need and scope of managerial direction. Using job posting data around ChatGPT's release, we find reduced managerial hiring, flatter reporting structures, and higher wages for non-managerial roles.

3. [Title hidden to ensure blind review -- a piece on Topic Modeling]

with: Simone Santoni & Vibha Gaba

Status: Under Review at Strategic Management Journal

Paper in a nutshell

This paper examines topic modeling as a method for large-scale text analysis in strategy research, focusing on the challenge of selecting the number of topics. We provide guidelines aligning research objectives, search strategies, and evaluation approaches, illustrated through an application on Generative AI's impact on managerial occupations.

4. Penguins Don't Talk, But They Huddle: How Plans Redistribute Feedback

with: Simone Santoni & Gianvito Lanzolla

Status: Working Paper

Job Market Paper

Abstract

Organizations coordinate activities among interdependent individuals using two broad mechanisms: feedback and organizational plans. While plans reduce the need for feedback by sorting and sequencing organizational actions, they also enhance feedback capacity by liberating cognitive resources. This apparent contradiction raises a critical question: how do organizations leverage feedback when plans are available? We argue that plans expose the means-ends chain connecting individual effort to rewards and theorize that the introduction and failure of plans reveal distinct effort-reward structures that shape feedback patterns. To test our theory, we analyze a 20-year dataset of Linux Kernel development, examining the effects of the 2005 plan introduction and subsequent instances of plan failure. Our results demonstrate that following plan introduction, organizations can coordinate more actions while requiring less communication per action. Conversely, cognitive resources become depleted when plans fail, as organizations must address more actions with higher communication volume. Additionally, we find that redistributing freed cognitive resources depends on the epistemic interdependence of activities.

5. CTRL-OSS: controlling Linux

with: Simone Santoni

Status: Working Paper

Abstract

This study considers the problem of control in open-source software. In particular, it aims to characterize control manifestations happening through language in an organizational setting where no formal authority exists. Hence, it focuses on how natural language turns into a control device in open-source software. To do so, it leverages naturally-occurring data from the `Linux Kernel' project, analyzed by integrating a qualitative approach to the study of meanings with a text classifier based on Deep Learning. Hence, it charts the topology of control as situated in peer-to-peer interaction, uncovering the individual facets of control and appreciating their relationship with linguistic means offered by online sociality. The topology results in a set of foci of control, linguistic drivers, and their association which is further tested against the community-level goals of participation and contribution. Overall, the project enquires about the boundaries and nature of control, offering momentum on how independent or organizational agents enforce control within organizations whose actors are not bound by any employment relationships and pursue different goals.

6. Leading DAOs: Between Formal Structure and Legitimate Authority

with: Oliver Alexy & Ying-Ying Hsieh

Status: Working Paper

Abstract

This research investigates legitimate authority in flat organizations. In particular, it aims to unpack what gives individuals the legitimate authority to influence the division of labor. Hence, it problematizes the role of formal structure -- i.e., formal leadership appointment -- vis-à-vis other established behavioral and structural characteristics for individuals to exert influence over tasks and agents. We study the above by leveraging a longitudinal dataset covering all peer-to-peer interactions in 10 Decentralized Autonomous Organizations. Our findings reveal the persistence of formal superior-subordinate relationships even in such forms of organizing, with formal leaders exerting greater influence than an average member. Yet, not all members are born equal. Thus, if compared to similar contributors, a formally appointed leader pays a penalty, suggesting that structure is not inherently consequential for legitimate authority. Our study highlights the limits of formally derived authority in flat organizations, demonstrating that while formal roles may facilitate managerial action, they must be legitimized through social processes to exert legitimate authority. Otherwise, the formal structure penalty propagates across behavioral and structural characteristics commonly associated with the ability to lead.

7. Give Me a Map: Qualitative Analysis of Large Textual Data with ML

with: Simone Santoni

Status: Working Paper

Abstract

Performing qualitative analysis on millions of textual documents presents a significant methodological challenge. As the data size grows, so does the risk of researchers getting trapped in 'concentration sites'—limited areas of insight that obscure a comprehensive understanding of the phenomenon, making the search for substantive meanings increasingly complex. While recent applications advocate for externalizing pattern discovery to Machine Learning (ML), this essay proposes an alternative: a methodological framework that combines traditional qualitative methods with supervised and unsupervised ML applications to enhance and scale human-led analysis. The framework leverages ML mapping and scaling capabilities to orient human attention and energy toward substantive semantic instances, allowing for discovering, refining, and assessing patterns.