Ordinal variables, such as on the Likert scale, are common in applied research. Yet, existing methods for causal inference tend to target nominal or continuous data. When applied to ordinal data, this fails to account for the inherent ordering or imposes well-defined relative magnitudes. Hence, there is a need for specialised methods to compute interventional effects between ordinal variables while accounting for their ordinality. One potential framework is to presume a latent Gaussian Directed Acyclic Graph (DAG) model: that the ordinal variables originate from marginally discretizing a set of Gaussian variables whose latent covariance matrix is constrained to satisfy the conditional independencies inherent in a DAG. Conditioned on a given latent covariance matrix and discretisation thresholds, we derive a closed-form function for ordinal causal effects in terms of interventional distributions in the latent space. Our causal estimation combines naturally with algorithms to learn the latent DAG and its parameters, like the Ordinal Structural EM algorithm. Simulations demonstrate the applicability of the proposed approach in estimating ordinal causal effects both for known and unknown structures of the latent graph. As an illustration of a real-world use case, the method is applied to survey data of 408 patients from a study on the functional relationships between symptoms of obsessive-compulsive disorder and depression.

We had quite an extensive discussion about this paper after the author response. The reviewers appreciated the clarifications and especially the additional experiment. What makes the paper stand out is proposing two generic conditions that enforce the emergence of spatial structures and experimentally validating them. The discussion circled around Equation (1), how well that would hold for realistic (noisy) sensing and what the implication on the emergence of the spatial structures would be. To our understanding the later parts of the paper don't rely on this equation but only on the intuition.

We focus this analysis on the United States' Department of Defense (DoD), where the US Army alone is projected to spend an estimated $5 billion per year (in 2020 dollar terms through 2050), developing and acquiring ground vehicles, where ground vehicles are any vehicles other than aircraft and ships (CBO 2021). Maintaining this enormous investment is critical to ensuring combat readiness across the DoD, where the department spent $90 billion in 2022 on maintaining vehicles across domains: ground, air, and sea (GAO 2022). Predicting requirements is critical to an effective maintenance program. The application of statistics towards vehicle maintenance prediction is often referred to as predictive maintenance. Recognizing the importance of predictive maintenance, in the 2022 National Defense Authorization Act (NDAA) Congress required the DoD Inspector General Office to review predictive maintenance practices, originally established by DoD directives in 2002 and 2007 (DoDIG 2023).

The author's recent research papers, "Cumulative deviation of a subpopulation from the full population" and "A graphical method of cumulative differences between two subpopulations" (both published in volume 8 of Springer's open-access "Journal of Big Data" during 2021), propose graphical methods and summary statistics, without extensively calibrating formal significance tests. The summary metrics and methods can measure the calibration of probabilistic predictions and can assess differences in responses between a subpopulation and the full population while controlling for a covariate or score via conditioning on it. These recently published papers construct significance tests based on the scalar summary statistics, but only sketch how to calibrate the attained significance levels (also known as "P-values") for the tests. The present article reviews and synthesizes work spanning many decades in order to detail how to calibrate the P-values. The present paper presents computationally efficient, easily implemented numerical methods for evaluating properly calibrated P-values, together with rigorous mathematical proofs guaranteeing their accuracy, and illustrates and validates the methods with open-source software and numerical examples.

What's coming: Elon Musk and Jeff Bezos are about to lock horns once again. Last month, the US Federal Communications Commission approved the final aspects of Project Kuiper, Amazon's effort to deliver high-speed internet access from space. In May, the company will test its satellites in an effort to take on SpaceX's own venture, Starlink, and tap into a potentially very lucrative market. The catch: The key difference is that Starlink is operational, and has been for years, whereas Amazon doesn't plan to start offering Kuiper as a service until 2024, giving SpaceX a considerable head start. Also, none of the rockets Amazon has bought a ride on has yet made it to space.

Human psychology may prevent people from realizing the benefits of artificial intelligence, according to a trio of boffins based in the Netherlands. But with training, we can learn to overcome our biases and trust our automated advisors. In a preprint paper titled "Knowing About Knowing: An Illusion of Human Competence Can Hinder Appropriate Reliance on AI Systems," Gaole He, Lucie Kuiper, and Ujwal Gadiraju, from Delft University of Technology, examine whether the Dunning-Kruger effect hinders people from relying on recommendations from AI systems. The Dunning-Kruger effect (DKE) dates back to research from 1999 by psychologists David Dunning and Justin Kruger, "Unskilled and unaware of it: How difficulties in recognizing one's own incompetence lead to inflated self-assessments." Dunning and Kruger posit that incompetent people lack the capacity to recognize their incompetence and thus tend to overestimate their abilities.

Self-driving cars and robots that can zoom on their own around warehouses are just some of what's possible because of artificial intelligence. But expect unforeseen consequences if researchers ignore the inherent ethical dilemmas in the emerging technology. That's one of the takeaways from a panel about AI ethics and education in San Francisco that was hosted by the Future of Life Institute, a research group focused on preventing societal problems created by the technology. Although humans typically program AI-powered robots to accomplish a particular goal, these robots will typically make decisions on their own to reach the goal, explained Benjamin Kuipers, a computer science professor and AI researcher at the University of Michigan. Get Data Sheet, Fortune's technology newsletter.

At the risk of being scolded again for "employing universal truths and unarguable facts" in support of my position, I must point out that it is the responsibility of a scientist or engineer to document clearly the known limitations of any method he develops and publishes. In addition to truth in packaging, a clear and unblinking examination of the limitations of one's own work is an invaluable guide to further research. Akman observes, correctly, that QSIM is a purely mathematical formalism for expressing qualitative differential equation models of the world, and not a physical modeling methodology. Our research group has also been concerned with this limitation, so we have developed modelbuilding methods which compile QDEs for QSIM to simulate, either from a component-connection description of a device (Franke and Dvorak 1989, 1990), or from a physical scenario description via qualitative views and processes (Crawford, Farquhar, and Kuipers 1990). These two model-building methods are important elements of the QSIM perspective on qualitative reasoning (Kuipers 1989).

Jim Saveland Research Forester Associate Editor, AI Application in Natural Resource Management United States Department of Agriculture Forest Service Southern Forest Fire Laboratory Route 1, Box 182A Dry Branch, GA 31020 Editor: Mr. Saveland's letter focuses our attention on the important distinction between accuracy and realism. We believed the Phoenix fire simulator to be accurate (with the provisos noted in our article). Mr. Saveland believes otherwise, and he is certainly better qualified than us to judge! We can allay some doubts (e.g., firefighting objects actually do move at variable rates, depending on ground cover, as Mr. Saveland notes they should), but basically we agree with Mr. Saveland that the Phoenix fire simulator is not accurate. But we do claim it is realistic.

We examine different formalisms for modeling qualitatively physical systems and their associated inferential processes that allow us to derive qualitative predictions from the models. We highlight the mathematical aspects of these processes along with their potential and limitations. However, the modeling process encounters difficulties from both ends: A model must adapt to the knowledge available and the task it is built for. The possible limitations of traditional numeric methods with respect to these problems mean qualitative models can be a good alternative: (1) qualitative models cope with uncertain and incomplete knowledge, (2) a qualitative model output equals an infinity of numeric runs that are obtained at once in compact form, (3) the qualitative predictions provide the relevant qualitative distinctions in the system's behavior, and (4) the modeling primitives allow for a more intuitive interpretation. A system's evolution can be tackled in discrete terms by defining states and events that trigger transitions between states.