Veranstaltungen

Titel und Abstract folgen.

Ansprechpartnerin: Elisabeth Hofmeister

Eintragung in den Einladungsverteiler und mehr Informationen auf der Seminarseite.

Titel und Abstract folgen.

Ansprechpartner: Dominik Asam

Eintragung in den Einladungsverteiler und mehr Informationen auf der Seminarseite.

Titel und Abstract folgen.

Ansprechpartner: Daehyun Kim

Eintragung in den Einladungsverteiler und mehr Informationen auf der Seminarseite.

Titel und Abstract folgen.

Ansprechpartner: Elisabeth Hofmeister

Eintragung in den Einladungsverteiler und mehr Informationen auf der Seminarseite.

Titel und Abstract folgen.

Ansprechpartnerin: Ulrike Morgalla

Eintragung in den Einladungsverteiler und mehr Informationen auf der Seminarseite.

Causal machine learning (Causal ML) is an emerging branch in ML/AI research aimed data-driven decision-making by integrating robust causal inference with advanced predictive algorithms. A key advantage of Causal ML is the ability to prediction under intervention, that is, to predict the outcomes of a treatment at the individualized level while adjusting for various confounders. Causal ML can explicitly model how the treatment impact varies across subpopulations, thus uncovering rich, individual-level heterogeneity that can be leveraged for personalized targeting and more effective decisions. In this talk, we explore the methodological foundations of Causal ML, discuss critical guardrails necessary for its rigorous and responsible deployment, and explore applications in behavioral science and policy. In particular, we introduce the “AI Heterogeneity Explorer”, which allows to uncover the differential effectiveness of behavioral interventions and thus identify for whom interventions are effective. The “AI Heterogeneity Explorer” provides a systematic recipe for understanding the heterogeneity of behavioral interventions, optimizing the personalized delivery of interventions, validating the targeting strategy—which offers a powerful alternative to one-size-fits-all approaches often used in data-driven decision-making. Finally, we illustrate how this explorer can be leveraged in the context of climate interventions to advance behavioral and climate science.

Ansprechpartner: Malte Toetzke

Eintragung in den Einladungsverteiler und mehr Informationen auf der Seminarseite.

Titel und Abstract folgen.

Ansprechpartner: Daehyun Kim

Eintragung in den Einladungsverteiler und mehr Informationen auf der Seminarseite.

Titel und Abstract folgen.

Ansprechpartnerin: Elisabeth Hofmeister

Eintragung in den Einladungsverteiler und mehr Informationen auf der Seminarseite.

Like physical products, new technologies are developed using globally sourced inputs. But while the supply chains behind physical goods are well understood, we know far less about the international ”supply chain” of scientifi c knowledge that powers U.S. innovation—and how vulnerable it may be to disruption. Here, I uncover the structure of the U.S. knowledge supply chain by tracing multi-generational citation paths that connect NSF-funded research to downstream patents, and assess its fragility by simulating barriers that impede the fl ow of scientifi c knowledge across the U.S. border. The results reveal that U.S. innovation is deeply reliant on foreign science: 56% of the intermediary papers linking NSF research to patents are produced outside the United States. Cross-border restrictions reduce the connectivity of these paths, increase their length, and lower innovation productivity, as measured by the U.S. patent-to-publication ratio. Most consequentially, such restrictions strand promising knowledge trajectories outside the U.S.: I estimate there are 104,149 NSF-stimulated paths currently under development outside the U.S. Under the status quo, 67,965 are projected to return to the U.S. for patenting. However, under scientifi c autarky, virtually none would, representing a loss of approximately $10.7 billion in invested capital. These impacts also affect U.S. fi rms that are critical to national priorities, including innovation, energy, and security. For example, autarky reduces outstanding path capture at Microsoft, ExxonMobil, and Lockheed Martin by between 48% and 57%.

Ansprechpartner: Daehyun Kim

Eintragung in den Einladungsverteiler und mehr Informationen auf der Seminarseite.

Generative AI and Community Norms in a User-Generated Media Commons: Evidence from Archive of Our Own
Presenter: Denzel Glandel (ISTO)
Discussant: Christian Untch (TUM)

Advances in generative artificial intelligence (GenAI) enable the large-scale production of creative content, raising questions about how algorithmically assisted works are received in community-governed digital markets. Prior research finds negative reactions to algorithm-created products, but it remains unclear how such reactions operate when production and evaluation are embedded in strong social-norm environments. I study Archive of Our Own (AO3), a fanfiction platform characterized by a gift economy and voluntary disclosure of GenAI use. Using story-level and author-level data on reader attention and feedback, I distinguish between exposure and conditional evaluation. The results show that GenAI disclosure has little effect on exposure but is associated with lower appreciation and reduced conversational engagement. These effects depend on the type of GenAI usage, and are stronger in fandoms with higher baseline feedback intensity. The findings extend research on algorithm valuation by highlighting the role of informal community governance in the responses to GenAI-produced content.

Access to the Frontier: Open Source AI and Downstream Innovation
Presenter: Dominik Asam (MPI)
Discussant: Svenja Friess (ISTO)
 

How does open sourcing a frontier AI model affect downstream innovation? I provide empirical evidence on the effect of shifting from a closed to an open frontier on subsequent AI development by studying the leak of Meta’s Llama model in early 2023. I find that Llama lowered entry barriers and led to a massive surge in new models without compromising average perceived utility. The increase in model creation is almost entirely driven by new entrants, who show similar experience characteristics to incumbents. This suggests that the previous dominance of closed models imposed technological entry barriers that left a large pool of skilled contributors untapped. An analysis of incumbents’ reactions to Llama shows that ex ante exposed contributors pivot away from Llama-affected domains towards unaffected fields, indicating a strategic reallocation of innovative effort. Overall, the results emphasize the importance of open source AI as a platform for experimentation and innovation.

Why Is It Still So Difficult? – A Multilevel Analysis Of Barriers To Commercializing Radical Innovations In Large Companies.
Presenter: Tim Hahne (TUM)
Discussant: Johannes Könemann (MPI)

Although widely recognized as being of fundamental importance, practice shows that especially large firms still face significant challenges in developing radical innovations in systematic and reliable ways. Particularly the commercialization phase constitutes substantial difficulties. This contradicts previous research, which has established such commercialization as relatively easy task for large firms due to high levels of resource and network availability. To investigate this phenomenon, we conducted a qualitative case study with our case company being a multinational top-player in the industrial IT sector. In total, we combine data from 14 months of participant observation, 32 semi-structured interviews and a wide variety of archival data. As a result, we propose a model of the three fundamental barriers to commercialization success of radical innovations in large, incumbent companies, and how these barriers interact: First, separation during exploration and incubation is beneficial. Yet, firms struggle with separate commercialization efforts, leaving them with the essential challenge of reintegration into their established business. Second, even if reintegrated, the potential to commercialize radical innovations is limited, if the distribution imbalance of commercialization resources between achieving short-term goals and long-term development of innovations remains unresolved. Finally, too rigid organizational structures prevent the exploration of required new commercialization strategies.

Ansprechpartnerin: Elisabeth Hofmeister