The formalization of action and obligation using logic languages is a topic of increasing relevance in the field of ethics for AI. Having an expressive syntactic and semantic framework to reason about agents decisions in moral situations allows for u
nequivocal representations of components of behavior that are relevant when assigning blame (or praise) of outcomes to said agents. Two very important components of behavior in this respect are belief and belief-based action. In this work we present a logic of doxastic oughts by extending epistemic deontic stit theory with beliefs. On one hand, the semantics for formulas involving belief operators is based on probability measures. On the other, the semantics for doxastic oughts relies on a notion of optimality, and the underlying choice rule is maximization of expected utility. We introduce an axiom system for the resulting logic, and we address its soundness, completeness, and decidability results. These results are significant in the line of research that intends to use proof systems of epistemic, doxastic, and deontic logics to help in the testing of ethical behavior of AI through theorem-proving and model-checking.
Most adversarial attack methods that are designed to deceive a text classifier change the text classifiers prediction by modifying a few words or characters. Few try to attack classifiers by rewriting a whole sentence, due to the difficulties inheren
t in sentence-level rephrasing as well as the problem of setting the criteria for legitimate rewriting. In this paper, we explore the problem of creating adversarial examples with sentence-level rewriting. We design a new sampling method, named ParaphraseSampler, to efficiently rewrite the original sentence in multiple ways. Then we propose a new criteria for modification, called a sentence-level threaten model. This criteria allows for both word- and sentence-level changes, and can be adjusted independently in two dimensions: semantic similarity and grammatical quality. Experimental results show that many of these rewritten sentences are misclassified by the classifier. On all 6 datasets, our ParaphraseSampler achieves a better attack success rate than our baseline.
In many real life situations, including job and loan applications, gatekeepers must make justified and fair real-time decisions about a persons fitness for a particular opportunity. In this paper, we aim to accomplish approximate group fairness in an
online stochastic decision-making process, where the fairness metric we consider is equalized odds. Our work follows from the classical learning-from-experts scheme, assuming a finite set of classifiers (human experts, rules, options, etc) that cannot be modified. We run separate instances of the algorithm for each label class as well as sensitive groups, where the probability of choosing each instance is optimized for both fairness and regret. Our theoretical results show that approximately equalized odds can be achieved without sacrificing much regret. We also demonstrate the performance of the algorithm on real data sets commonly used by the fairness community.
The Kapteyn moving group has been postulated as tidal debris from $omega$ Centauri. If true, members of the group should show some of the chemical abundance patterns known for stars in the cluster. We present an optical and near-infrared high-resolut
ion, high-S/N spectroscopic study of 14 stars of the Kapteyn group, plus 10 additional stars (the $omega$ Cen-group) that, while not listed as members of the Kapteyn group as originally defined, have been nevertheless associated dynamically with $omega$ Centauri. Abundances for Na, O, Mg, Al, Ca and Ba were derived from the optical spectra, while the strength of the chromospheric He I 10830 {AA} line is studied as a possible helium abundance indicator. The resulting Na-O and Mg-Al patterns for stars of the combined Kapteyn and $omega$ Cen-group samples do not resemble those of $omega$ Centauri, and are not different from those of field stars of the Galactic halo. The distribution of equivalent widths of the He I 10830 {AA} line is consistent with that found among non-active field stars. Therefore, no evidence is found for second-generation stars within our samples, which most likely rules out a globular-cluster origin. Moreover, no hint of the unique Ba-overabundance at the metal-rich end, well-established for $omega$ Centauri stars, is seen among stars of the combined samples. Because this specific Ba pattern is present in $omega$ Centauri irrespective of stellar generation, this would rule out the possibility that our entire sample might be composed of only first generation stars from the cluster. Finally, for the stars of the Kapteyn group, the possibility of an origin in the hypothetical $omega$ Centauris parent galaxy is disfavored by the different run of $alpha$-elements with metallicity between our targets and stars from present-day dwarf galaxies.
We present the first detailed chemical abundance analysis of the old 8.2 Gyr solar twin, HIP 102152. We derive differential abundances of 21 elements relative to the Sun with precisions as high as 0.004 dex ($lesssim$1%), using ultra high-resolution
(R = 110,000), high S/N UVES spectra obtained on the 8.2-m Very Large Telescope. Our determined metallicity of HIP 102152 is [Fe/H] = -0.013 $pm$ 0.004. The atmospheric parameters of the star were determined to be 54 K cooler than the Sun, 0.09 dex lower in surface gravity, and a microturbulence identical to our derived solar value. Elemental abundance ratios examined vs. dust condensation temperature reveal a solar abundance pattern for this star, in contrast to most solar twins. The abundance pattern of HIP 02152 appears to be the most similar to solar of any known solar twin. Abundances of the younger, 2.9 Gyr solar twin, 18 Sco, were also determined from UVES spectra to serve as a comparison for HIP 102152. The solar chemical pattern of HIP 102152 makes it a potential candidate to host terrestrial planets, which is reinforced by the lack of giant planets in its terrestrial planet region. The following non-local thermodynamic equilibrium Li abundances were obtained for HIP 102152, 18 Sco, and the Sun: log $epsilon$ (Li) = 0.48 $pm$ 0.07, 1.62 $pm$ 0.02, and 1.07 $pm$ 0.02, respectively. The Li abundance of HIP 102152 is the lowest reported to date for a solar twin, and allows us to consider an emerging, tightly constrained Li-age trend for solar twin stars.
