Search

Counterfactual explanations are increasingly used as an Explainable Artificial Intelligence (XAI) technique to provide stakeholders of complex machine learning algorithms with explanations for data-driven decisions. The popularity of counterfactual explanations resulted in a boom in the algorithms generating them. However, not every algorithm creates uniform explanations for the same instance. Even though in some contexts multiple possible explanations are beneficial, there are circumstances where diversity amongst counterfactual explanations results in a potential disagreement problem among stakeholders. Ethical issues arise when for example, malicious agents use this diversity to fairwash an unfair machine learning model by hiding sensitive features. As legislators worldwide tend to start including the right to explanations for data-driven, high-stakes decisions in their policies, these ethical issues should be understood and addressed. Our literature review on the disagreement problem in XAI reveals that this problem has never been empirically assessed for counterfactual explanations. Therefore, in this work, we conduct a large-scale empirical analysis, on 40 datasets, using 12 explanation-generating methods, for two black-box models, yielding over 192.0000 explanations. Our study finds alarmingly high disagreement levels between the methods tested. A malicious user is able to both exclude and include desired features when multiple counterfactual explanations are available. This disagreement seems to be driven mainly by the dataset characteristics and the type of counterfactual algorithm. XAI centers on the transparency of algorithmic decision-making, but our analysis advocates for transparency about this self-proclaimed transparency

The Event Horizon Telescope (EHT) has led to the first images of a supermassive black hole, revealing the central compact objects in the elliptical galaxy M87 and the Milky Way. Proposed upgrades to this array through the next-generation EHT (ngEHT) program would sharply improve the angular resolution, dynamic range, and temporal coverage of the existing EHT observations. These improvements will uniquely enable a wealth of transformative new discoveries related to black hole science, extending from event-horizon-scale studies of strong gravity to studies of explosive transients to the cosmological growth and influence of supermassive black holes. Here, we present the key science goals for the ngEHT and their associated instrument requirements, both of which have been formulated through a multi-year international effort involving hundreds of scientists worldwide., Comment: 32 pages, 11 figures, accepted for publication in a special issue of Galaxies on the ngEHT (https://www.mdpi.com/journal/galaxies/special_issues/ngEHT_blackholes)

Multiple viable theoretical models predict heavy dark matter particles with a mass close to the Planck mass, a range relatively unexplored by current experimental measurements. We use 219.4 days of data collected with the XENON1T experiment to conduct a blind search for signals from Multiply-Interacting Massive Particles (MIMPs). Their unique track signature allows a targeted analysis with only 0.05 expected background events from muons. Following unblinding, we observe no signal candidate events. This work places strong constraints on spin-independent interactions of dark matter particles with a mass between 1$\times$10$^{12}\,$GeV/c$^2$ and 2$\times$10$^{17}\,$GeV/c$^2$. In addition, we present the first exclusion limits on spin-dependent MIMP-neutron and MIMP-proton cross-sections for dark matter particles with masses close to the Planck scale., Comment: 7 pages, 6 figures

Within its Voyage 2050 planning cycle, the European Space Agency (ESA) is considering long-term large class science mission themes. Gravitational-wave astronomy is among the topics under study. Building on previous work by other authors, this paper studies a gravitational-wave interferometer concept, dubbed "LISAmax", consisting of three spacecraft, each located close to one of the Sun-Earth libration points L3, L4 and L5, forming a triangular constellation with an arm length of 259 million kilometers (to be compared to LISA's 2.5 million kilometer arms). We argue that this is the largest triangular formation that can be reached from Earth without a major leap in mission complexity and cost (hence the name). The sensitivity curve of such a detector is at least two orders of magnitude lower in amplitude than that of LISA, at frequencies below 1 mHz. This makes the observatory sensitive to gravitational waves in the {\mu}Hz range and opens a new window for gravitational-wave astronomy, not covered by any other planned detector concept. We analyze in detail the constellation stability for a 10-year mission in the full numerical model including insertion, dispersion, and self-gravity-induced accelerations. We compute the orbit transfers using a European launcher and chemical propulsion. Different orbit options, such as precessing, inclined orbits, the use of flybys for the transfer, and the launch strategy, are discussed. The payload design parameters are assessed, and the expected sensitivity curve is compared with a number of potential gravitational-wave sources. No show stoppers are identified at this point of the analysis., Comment: 20 pages, 11 figures, 3 tables

We developed a detector signal characterization model based on a Bayesian network trained on the waveform attributes generated by a dual-phase xenon time projection chamber. By performing inference on the model, we produced a quantitative metric of signal characterization and demonstrate that this metric can be used to determine whether a detector signal is sourced from a scintillation or an ionization process. We describe the method and its performance on electronic-recoil (ER) data taken during the first science run of the XENONnT dark matter experiment. We demonstrate the first use of a Bayesian network in a waveform-based analysis of detector signals. This method resulted in a 3% increase in ER event-selection efficiency with a simultaneously effective rejection of events outside of the region of interest. The findings of this analysis are consistent with the previous analysis from XENONnT, namely a background-only fit of the ER data., Comment: 11 pages, 8 figures

This white paper outlines the plans of the History Philosophy Culture Working Group of the Next Generation Event Horizon Telescope Collaboration., Comment: 23 pages, 1 figure

Algorithms used by organizations increasingly wield power in society as they decide the allocation of key resources and basic goods. In order to promote fairer, juster, and more transparent uses of such decision-making power, explainable artificial intelligence (XAI) aims to provide insights into the logic of algorithmic decision-making. Despite much research on the topic, consumer-facing applications of XAI remain rare. A central reason may be that a viable platform-based monetization strategy for this new technology has yet to be found. We introduce and describe a novel monetization strategy for fusing algorithmic explanations with programmatic advertising via an explanation platform. We claim the explanation platform represents a new, socially-impactful, and profitable form of human-algorithm interaction and estimate its potential for revenue generation in the high-risk domains of finance, hiring, and education. We then consider possible undesirable and unintended effects of monetizing XAI and simulate these scenarios using real-world credit lending data. Ultimately, we argue that monetizing XAI may be a double-edged sword: while monetization may incentivize industry adoption of XAI in a variety of consumer applications, it may also conflict with the original legal and ethical justifications for developing XAI. We conclude by discussing whether there may be ways to responsibly and democratically harness the potential of monetized XAI to provide greater consumer access to algorithmic explanations.

We report on the first search for nuclear recoils from dark matter in the form of weakly interacting massive particles (WIMPs) with the XENONnT experiment which is based on a two-phase time projection chamber with a sensitive liquid xenon mass of $5.9$ t. During the approximately 1.1 tonne-year exposure used for this search, the intrinsic $^{85}$Kr and $^{222}$Rn concentrations in the liquid target were reduced to unprecedentedly low levels, giving an electronic recoil background rate of $(15.8\pm1.3)~\mathrm{events}/(\mathrm{t\cdot y \cdot keV})$ in the region of interest. A blind analysis of nuclear recoil events with energies between $3.3$ keV and $60.5$ keV finds no significant excess. This leads to a minimum upper limit on the spin-independent WIMP-nucleon cross section of $2.58\times 10^{-47}~\mathrm{cm}^2$ for a WIMP mass of $28~\mathrm{GeV}/c^2$ at $90\%$ confidence level. Limits for spin-dependent interactions are also provided. Both the limit and the sensitivity for the full range of WIMP masses analyzed here improve on previous results obtained with the XENON1T experiment for the same exposure., Comment: Limit points are included in the submission file

We report on the first simultaneous measurement of high-order harmonics, relativistic electrons and low divergence proton beams generated from plasma mirrors driven at kHz repetition rate by relativistic-intensity milliJoule-energy femtosecond laser pulses. This setup enables detailed parametric studies of the particle and radiation spatio-spectral beam properties for a wide range of controlled interaction conditions, such as pulse duration and plasma density scale length. This versatile setup should aid in further understanding the collective laser absorption mechanisms at play during the laser-plasma interaction and in optimizing the secondary beam properties for potential applications.

Yield stress materials fail when the imposed stress crosses a critical threshold. A well-known dynamical response to the applied stress is the phenomenon of creep where the cumulative deformation grows sublinearly with time, prior to failure or arrest. Using extensive molecular dynamics simulations, we study such response for a model amorphous system, in the athermal limit, and probe how the annealing history of the initial state determines the observed behaviour to an applied shear stress. Further, we analyze the microscopic dynamics in the vicinity of the yield threshold, using large systems, and characterize the spatiotemporal signatures towards arrest or flow, at different scales.

We study both numerically and experimentally the use of two third-order nonlinear temporal filtering techniques, namely nonlinear ellipse rotation (NER) and cross-polarized wave (XPW) generation, for spatio-temporal cleaning of mJ energy 30 fs Titanium:Sapphire laser pulses in a multi-pass cell. In both cases, a contrast enhancement greater than 3 orders of magnitude is observed, together with excellent output pulse quality and record high conversion efficiencies. Careful balancing of nonlinearity and dispersion inside the multi-pass cell helps tune the spectral broadening process and control the post-compressed pulse duration for specific applications.

Skip connections and normalisation layers form two standard architectural components that are ubiquitous for the training of Deep Neural Networks (DNNs), but whose precise roles are poorly understood. Recent approaches such as Deep Kernel Shaping have made progress towards reducing our reliance on them, using insights from wide NN kernel theory to improve signal propagation in vanilla DNNs (which we define as networks without skips or normalisation). However, these approaches are incompatible with the self-attention layers present in transformers, whose kernels are intrinsically more complicated to analyse and control. And so the question remains: is it possible to train deep vanilla transformers? We answer this question in the affirmative by designing several approaches that use combinations of parameter initialisations, bias matrices and location-dependent rescaling to achieve faithful signal propagation in vanilla transformers. Our methods address various intricacies specific to signal propagation in transformers, including the interaction with positional encoding and causal masking. In experiments on WikiText-103 and C4, our approaches enable deep transformers without normalisation to train at speeds matching their standard counterparts, and deep vanilla transformers to reach the same performance as standard ones after about 5 times more iterations., Comment: ICLR 2023

We examine the structural relaxation of glassy materials at finite temperatures, considering the effect of activated rearrangements and long-range elastic interactions. Our three-dimensional mesoscopic relaxation model shows how the displacements induced by localized relaxation events can result in faster-than-exponential relaxation. Thermal activation allows for local rearrangements, which generate elastic responses and possibly cascades of new relaxation events. To study the interplay between this elastically-dominated and thermally-dominated dynamics, we introduce tracer particles that follow the displacement field induced by the local relaxation events and also incorporate Brownian motion. Our results reveal that the dynamic exponents and shape parameter of the dynamical structure factor depend on this competition and display a crossover from faster-than-exponential to exponential relaxation as temperature increases, consistent with recent observations in metallic glasses. Additionally, we find the distribution of waiting times between activations to be broadly distributed at low temperatures, providing a measure of dynamical heterogeneities characteristic for to glassy dynamics., Comment: 20 pages, 19 figures