Archivio istituzionale della Ricerca - Bocconi
Not a member yet
30225 research outputs found
Sort by
Prefazione a A. Franzoso, La Costituzione allo specchio. I principi fondamentali e le sfide del mondo di oggi
No full textIl contributo propone una lettura della Costituzione come patto di cittadinanza vivo e attuale, capace di orientare le scelte quotidiane e le trasformazioni sociali. Attraverso il ricorso alle storie concrete, i principi costituzionali emergono come criteri di giudizio e di responsabilità condivisa. La Costituzione è presentata non come testo astratto, ma come esperienza vissuta, che unisce diritti e doveri e chiede una partecipazione consapevole e continua da parte dei cittadin
Opportunities and Threats for Brands in the Social Media Ecosystem
No full textIn today’s digital environment, social media has become a central channel for understanding
consumers’ opinions, preferences, and behaviors. These platforms enable individuals to connect
with others who share similar interests, which leads to repeated exposure to brand-related content
within their personal networks. This can benefit brands by strengthening engagement and
fostering customer loyalty. At the same time, the same mechanisms that support rapid
information sharing can also facilitate the spread of misinformation, distort public perception,
and increase reputational risk. In parallel, recent advances in generative artificial intelligence
offer new tools for anticipating consumer responses, particularly in contexts where direct
feedback is limited or delayed. This thesis examines these three distinct but related areas, social
media echo chambers, misinformation, and generative AI, through a series of three essays.
The first essay, “Brand Echo Chambers,” extends the concept of echo chambers beyond
political discourse to explore their relevance in the context of brands. It investigates how
consumers who are surrounded by others with similar brand preferences perceive and engage
with brands. Drawing on data from over 870,000 users following 179 brands on X (formerly
Twitter), I introduce two distinct metrics to assess brand echo chambers. The findings show that
consumers in stronger brand echo chambers are more likely to discuss the brand, express positive
sentiment, and display emotional attachment, while being less exposed to competing brands.
The second essay, “The Cost of Misinformation for Brands: The Case of Bonduelle,” examines
misinformation targeting brands in digital environments, the role of automated accounts in
amplifying false claims, and the resulting impact on brand performance. Drawing on social
media data from a real-world misinformation campaign, I show that bots play a significant role
in amplifying false narratives and that this amplification coincides with a measurable decline in
the targeted firm’s stock price. This study underscores the strategic risks that misinformation
poses to brands and the influence of bots in shaping public perception.
The third essay, “Voices from the Future: Generative AI for Forecasting the Success of Artistic
Works,” explores whether synthetic content produced by large language models can serve as a
proxy for user feedback prior to a product’s release. Focusing on cultural products (books,
music, and films) I develop a two-step generation pipeline that first extracts media features and
metadata of each product, then produces simulated user reviews. I evaluate the semantic and
sentiment similarity between AI-generated and real user feedback and demonstrate that features
derived from synthetic reviews can predict post-release user ratings. This essay presents a novel
application of generative AI for early-stage market forecasting, especially in situations where
traditional feedback is delayed or unavailable.
Together, these essays contribute to a broader understanding of how digital environments enable
insight into consumer perceptions and behaviors, facilitate the circulation of false information
against brands, and support the use of emerging technologies to anticipate customers’ feedback.
More broadly, it highlights the evolving challenges and opportunities that arise as firms and
consumers interact within increasingly complex and algorithmically mediated systems
Fostering creativity, innovation, and CSR through management control systems
No full textIn the first chapter, I explore how artificial intelligence (AI) evaluators of creative outputs, instead of human evaluators, influence employees’ creativity. I also investigate whether the effect depends on the type of feedback provided by the evaluator and on whether the evaluation influences employees’ rewards. In an online experiment, I find that AI evaluators reduce employees’ creativity compared with human evaluators when the feedback includes numeric scores and the evaluation influences employees’ rewards. In contrast, the effect is not significant when the feedback is non-numeric or the evaluation does not influence employees’ rewards. Employees perceive AI evaluators that generate numeric feedback as adopting a quantitative evaluation approach. This perception diminishes intrinsic motivation, thereby reducing employees’ creativity. Results can be explained by a process of perspective-taking. Employees may struggle to anticipate the assessment of evaluators that they perceive as different from themselves, such as AIs adopting a quantitative evaluation approach.
In the second chapter, coauthored with Eddy Cardinaels and Naomi Soderstrom, we study how mission statements interact with compensation interdependence to influence employees’ CSR engagement. When incentive plans tie employees’ wealth to financial performance, employees often face conflicts between prioritizing financial performance and pursuing prosocial outcomes (e.g., external CSR). This tension amplifies under compensation interdependence, where prosocial actions that reduce financial performance may also reduce colleagues’ pay. Employees may fear being viewed negatively by their peers for prosocial decisions that harm their company’s financial performance and mutual bonuses when their compensation is interdependent. In an online experiment, we document that stakeholder-oriented mission statements can mitigate this tension by fostering social norms that legitimize care for external stakeholders. Under compensation interdependence, these mission statements lead to more prosocial choices benefiting external stakeholders than shareholder-oriented missions. However, when compensation interdependence is absent, the type of mission statement does not alter employees’ decisions. Our research contributes to the literature on mission statements as management control systems, highlighting that their importance may increase with compensation interdependence.
In the third and final chapter, coauthored with Ariela Caglio and Angelo Ditillo, we examine the determinants and consequences of incentive plan modifications favoring CEOs (IPMs) after uncontrollable events that hinder target attainment. Using data from S&P 500 firms during 2020, the first year of COVID-19, we find that IPMs are more likely in firms that are more severely affected by the pandemic and when CEOs have stronger departure incentives. However, such IPMs are less likely in firms with a compensation committee, suggesting that such modifications are not always in the best interest of shareholders. Examining investment decisions, we find that IPMs are associated with increases in R&D spending and environmental and social (ES) performance. This association is more pronounced for IPMs affecting long-term incentive plans rather than short-term ones. Taken together, these results suggest that, while IPMs may raise governance concerns, they can also mitigate managerial myopia by encouraging longer-term investments
The relation between conditional conservatism and managerial herding: evidence from M&A waves
No full textWe examine the association between conditional conservatism and managerial herding in an M&A setting. Managerial herding occurs when managers imitate other firms' behaviour and rely less on private signals. In the M&A context, herding is more likely when stock markets are booming and managers underweight target fundamentals and private inputs to obtain short-term gains at the expense of long-term performance. We argue that a firm's commitment to conservatism reduces herding, making managers focus more on target fundamentals and private signals than outsiders' behaviour. Our empirical evidence confirms our hypothesis: we find that more conservative firms have a lower probability of undertaking M&As during booming markets. This lower probability, in turn, is associated with higher acquisition performance. By focusing on a distinct source of investment inefficiency, our results uncover an additional benefit of conditional conservatism thus contributing to the literature on the real effects of conservatism
A communist, an environmentalist and an android walk into a bar: the measurement and measurable effects of elite communication
No full textThere have never been as many things to talk about, or channels through which to talk about them, as there are today. The academic exercise of tracking, collecting and analyzing what is said is an established practice, but its form is in constant and accelerating mutation. In this dissertation, I empirically explore the impact of elite communication on two grand outcomes - national elections and asset pricing - with a strong methodological focus that explores different data collection and treatment procedures and incorporates recent developments in large language model (LLM) technology. I further design and describe an original method for audiovisual data collection that aims to greatly facilitate access to an extremely large and rich, but relatively hard to explore corpus of data, allowing for the extension of existing studies and undertaking of many new ones. By exploring and applying computational tools such as machine learning, LLMs or face and voice recognition, this work stresses the richness of political statement data and strives to demonstrate how to extract the most of and from it.
In the first chapter, through an original dataset of tweets by Portuguese politicians and a natural experiment derived from the outbreak of the Russo-Ukrainian war, I evaluate how stigmatizing behavior towards a radical left party can impact electoral results. Using regression discontinuity design and difference in differences approaches for inference, I obtain results that indicate that the stigmatized party suffers persistent vote intention losses.
The second chapter looks at a Twitter dataset covering German publicly traded firms and Swedish activist Greta Thunberg. It investigates how vocal activism by renowned opinion leaders can create an impact on firms' stock market performance, and how firm behavior might influence this relationship. Results suggest that companies that align themselves with opinion leaders can pass through this process unscathed, while others suffer a stock price decrease. In both the first and second chapter, machine learning and LLM classifiers are employed to refine the data by extracting meaning indicators from the raw text inputs.
The third chapter, finally, describes a method for building an analyzable transcript from audiovisual political data, such as broadcast debates or interviews, allowing for individual speaker diarization and recognition with minimal manual prep-work. This framework can be applied to material in any language, while its agile nature means it is easily adaptable to the specificities of different formats (e.g. short-form social media videos, multi-participant debates, live broadcasts). Through these avenues, it bears the potential to massively expand the amount of available data for political analysis.
This work makes several contributions. Firstly, it employs different methods for the collection and treatment of text, exemplifying their usage, allowing for their comparison, and using them for robust inference. Secondly, it approaches these exercises in a constructive way, aiming to provide better means of obtaining raw data and refining it into its most useful state. It shows, thus, how to employ LLM-based approaches to improve on mainstay methods such as machine learning classifiers or interpolation processes. Thirdly, it introduces a method that provides easily implementable and mostly-automatic access to a particularly rich type of data that was previously hidden behind either notoriously laborious or methodologically complex processes - debate and interview transcripts - and is ready to be adapted to alternative inputs. Insofar as political elites use the different channels at their disposal to convey a cohesive message, these approaches are likely to provide full coverage of politician stances and interventions. However, they go even deeper by tracking each individual to a level of granularity that easily allows for intra-politician message analysis
Learning on the Road: The MAMA Grand Tour as a Contemporary Field-Based Learning Model for Arts Managers
No full textThe article discusses the relevance of experiential learning through study tours and off campus visits in the context of arts management
Responses to Outcome Disclosure: People Asymmetrically Disclose or Hide Their Outcomes to Protect Others’ Emotions
No full textThis paper examines how what people disclose about their successes or failures depends on what others have disclosed. We propose that these decisions are guided less by self-focused motives and more by a concern for how one’s words will affect the other person’s emotions. Across nine studies (N = 8,229, including preregistered experiments, 2,216 self-written responses, and 473 real conversation dyads), we find that responders are consistently more likely to disclose matching outcomes (e.g., failures in response to failures) than non-matching ones (e.g., failures in response to successes), but with two asymmetries not predicted by prior theories. First, responders are more likely to disclose matching failures (failures in response to failures) than matching successes (successes in response to successes). Second, when experiencing non-matching outcomes, responders are more likely to disclose failures in response to successes than they are to disclose successes in response to failures. These patterns reflect other-focused attempts to comfort those who have failed and avoid exacerbating their distress. Beyond whether they disclosed, responders also adjusted how they disclosed, for instance, softening success disclosures in response to failures with consolation or apologies. These effects generalized across domains (e.g., health, career, financial), across relationships varying in closeness and status, and emerged in choices between pre-written responses, self-generated responses, and live conversations involving actual interpersonal disclosures. Disclosure decisions were moderated by factors such as liking and domain relevance. By demonstrating that responders’ outcome disclosures are systematically shaped by concern for the well-being of others, this work reframes disclosure as an intended conversational tool for protecting others’ emotions rather than managing self-presentation
The Foundations of Bayesian Entrepreneurship
No full textIntroductory chapter of the book that lays out the foundations of Bayesian entrepreneurshi
Algorithms Beyond the Union Bound: Polynomial Optimization and Discrepancy Theory
No full textThe union bound is a classical tool in the probabilistic method for proving the existence of objects with extremal features by showing that a random object satisfies each feature with high probability. This approach has powered major results spanning theoretical computer science, combinatorics, random matrix theory, and statistical physics — such as the existence of graph sparsifiers, satisfying assignments to constraint satisfaction problems, and low-energy configurations in spin glass models. However, the union bound ignores correlations between different features, and thereby often leads to suboptimal results in these applications.
This thesis presents an alternative approach: proving existence by designing and analyzing algorithms that construct the desired objects. We show that all the above problems can be formulated as minimizing either low-degree polynomials or linear-image norms in high dimensions. For these objectives, we analyze algorithms inspired by continuous optimization.
In the first part, we propose an orthogonal representation for first-order algorithms optimizing random quadratic polynomials using Fourier analysis. We show that in the high-dimensional limit, these algorithms can be analyzed by tracking their dynamics in a tree basis. This reframes random polynomial optimization as a combinatorial problem, which we explicitly solve in some cases. Our approach also yields the first direct justification in this setting of the heuristic cavity method from physics.
The second part extends the approach to general polynomial optimization. We introduce a multiscale union bound argument for random tensors, extending results of Friedman and Wigderson on the spectral gap of random hypergraphs. We further present a new rounding scheme for semidefinite programming relaxations, leading to improved approximations for homogeneous cubic optimization and the Max-3-Sat problem.
In the third part, we design an algorithm for minimizing norms of linear functions via Newton’s method applied to a regularized objective function. By varying the regularizer, we recover and generalize foundational results in discrepancy theory. This framework yields an improved bound in Spencer’s classical result from the 1980s, and establishes that the Beck–Fiala and Komlós conjectures hold for a new class of pseudorandom instances.
Together, these results show that algorithms can not only match, but also exceed the power of probabilistic arguments for finding extremal objects
