Talks

Many philosophers have argued that there is no causation in fundamental physics. On the other hand, physicists regularly use causal terminology and discuss the causal structure of physical theories. In this talk I will discuss a number of philosophical views on causation and try to understand how these differing points of view can be reconciled. I will argue that fundamental physics does indeed exhibit a structure that we might call ‘causal order,’ but this notion is weaker than the common-sense conception of causation because it does not involve asymmetry. I discuss the role played by causal order in fundamental physics and why it may be important to distinguish between various forms of causation, and I consider how the distinction between causal order and causation in the ordinary sense can help us better understand physical phenomena.

Title and abstract coming soon.

Process matrices are an effective tool to study causal relations across quantum system, with applications from foundational questions about causality to memory effects in open quantum systems. However, the current formulation assumes a discrete set of events at which systems can be probed, while prominent questions in quantum field theory, quantum gravity, and open-system dynamics are more natural in the continuum. I will present a continuous-time version of the framework for causally ordered processes (or combs) based on path integral formulation of a continuous Born rule, which cleanly separates processes from operations. The framework has direct applications to the study of continuous monitoring of open quantum systems with memory, overcoming limitations of previous approaches that require either a full microscopic model or strong Markovianity assumptions. It further allows extending to the continuum operational definitions of causality and Markovianity that characterise, respectively, processes compatible with a fixed causal order and processes without memory.

We’ll first discuss in this presentation causal discvory and reasoning in several causal abstractions, for both time series and standard data. We will in particular detail results on cluster DAGs with potential cycles, which corresponds to a setting in which partition admissibility is relaxed. We will then discuss the causal reasoning capacity of Large language Models (LLMs) and will present a recent work on leveraging  the predictive power of LLMs to infer causal relations between only two sequences of events.

The quantum switch is a prominent example of a process with an indefinite causal order that can be experimentally implemented. However, experimental implementations have faced criticisms, calling for a loophole free demonstration of indefinite causal order. Although it is still not known how to achieve this goal, in this talk I will present our recent experimental work in this regard. I will first provide a general overview of experimental implementations of the quantum switch. I will then discuss our experiment implementing a device-dependent test of indefinite causal order [1] and identify loopholes in our implementation. I will focus on loopholes related to the fact that measurement results are only read out at the end of the experiment. Following this, I will present our experimental implementation of a local measurement in the quantum switch [2], that bypasses this problem and show how it can be used for quantum key distribution in an indefinite causal order [3].

[1] C.M.D. Richter, M. Antesberger, H. Cao, P. Walther, L.A. Rozema, Towards an Experimental Device-Independent Verification of Indefinite Causal Order. PRX Quantum 7, 010354 (2026).
[2] Y. Valibouse, M. Cladera-Rosselló, M. Antesberger, P. Lima, P. Walther, L.A. Rozema. Time-Delocalized Local Measurements in an Indefinite Causal Order. arXiv:2604.11878.
[3] Hector Spencer-Wood. Indefinite causal key distribution J. Phys. A: Math. Theor. 58 495303 (2025).

The framework of process theories (a.k.a. Symmetric monoidal categories) has become quite ubiquitous in the description of physical theories, subsuming, for instance, generalised probabilistic theories and operational probabilistic theories. Process theories, however, come with quite a rigid notion of composition which causes tension when we start thinking about modifications to the standard notions of causality and causal structure. In recent work, arXiv:2502.10368, we have been developing a broader notion of process theory which, amongst other things, allows for a more flexible compositional, and hence causal, structure. In this talk I will introduce the basics of this framework, and how this relates to various attempts to modify causality in the literature.

Non-signalling conditions encode minimal requirements that any (quantum) systems must satisfy in order to be consistent with special relativity. Recent works have argued that in scenarios involving more than two parties, correlations compatible with relativistic causality do not have to satisfy all possible non-signalling conditions but only a subset of them. These are called relativistically causal correlations. Here we show that the set of relativistically causal correlations includes correlations that satisfy highly non-local monogamy relations between the effects of space-like separated random variables. These monogamy relations take the form of entropic inequalities and we give a general method to derive them. Using these monogamy relations, we exclude the possibility of certain physical mechanisms, for which it was previously known that they could explain specific examples of relativistically causal correlations. We achieve this by showing that there are setups in which such mechanisms in combination with the relaxed non-signalling conditions lead to superluminal signalling between space-like separated parties.

Closed timelike curves (CTCs) not being ruled out by the laws of physics has led quantum information scientists to formulate abstract mathematical models of their dynamical realization. Two leading models are Deutschian and postselected CTCs, the former introduced by David Deutsch in 1991 and the latter based on postselected teleportation. In the first model, in order to avoid paradoxes associated with CTCs, Deustch postulated that nature imposes a self-consistency condition on the density matrix of the chronology-violating qubits. In the second model, one avoids CTC paradoxes by postulating that the chronology-violating qubits are sent through the future mouth of a wormhole via a postselected teleportation protocol, whereby one always “gets lucky” in a teleportation protocol with only one particular measurement outcome occurring. Both models effect drastic changes to quantum mechanics and imply superior information-processing capabilities beyond what is believed to be reasonable in quantum information theory. In this talk, I will discuss both of the models and particular quantum circuit constructions that violate the uncertainty principle and no-cloning theorem of quantum mechanics. I will also discuss a surprising recent result that places an upper bound on how much information a sender and receiver can communicate over a noisy quantum channel assisted by post-selected closed timelike curves.