Search

Political polling is a multi-billion dollar industry with outsized influence on the societal trajectory of the United States and nations around the world. However, it has been challenged by factors that stress its cost, availability, and accuracy. At the same time, artificial intelligence (AI) chatbots have become compelling stand-ins for human behavior, powered by increasingly sophisticated large language models (LLMs). Could AI chatbots be an effective tool for anticipating public opinion on controversial issues to the extent that they could be used by campaigns, interest groups, and polling firms? We have developed a prompt engineering methodology for eliciting human-like survey responses from ChatGPT, which simulate the response to a policy question of a person described by a set of demographic factors, and produce both an ordinal numeric response score and a textual justification. We execute large scale experiments, querying for thousands of simulated responses at a cost far lower than human surveys. We compare simulated data to human issue polling data from the Cooperative Election Study (CES). We find that ChatGPT is effective at anticipating both the mean level and distribution of public opinion on a variety of policy issues such as abortion bans and approval of the US Supreme Court, particularly in their ideological breakdown (correlation typically >85%). However, it is less successful at anticipating demographic-level differences. Moreover, ChatGPT tends to overgeneralize to new policy issues that arose after its training data was collected, such as US support for involvement in the war in Ukraine. Our work has implications for our understanding of the strengths and limitations of the current generation of AI chatbots as virtual publics or online listening platforms, future directions for LLM development, and applications of AI tools to the political domain. (Abridged), Comment: 23 pages, 7 figures

What would the inputs be to a machine whose output is the destabilization of a robust democracy, or whose emanations could disrupt the political power of nations? In the recent essay "The Coming AI Hackers," Schneier (2021) proposed a future application of artificial intelligences to discover, manipulate, and exploit vulnerabilities of social, economic, and political systems at speeds far greater than humans' ability to recognize and respond to such threats. This work advances the concept by applying to it theory from machine learning, hypothesizing some possible "featurization" (input specification and transformation) frameworks for AI hacking. Focusing on the political domain, we develop graph and sequence data representations that would enable the application of a range of deep learning models to predict attributes and outcomes of political, particularly legislative, systems. We explore possible data models, datasets, predictive tasks, and actionable applications associated with each framework. We speculate about the likely practical impact and feasibility of such models, and conclude by discussing their ethical implications., Comment: 11 2-column pages, 2 figures

Automated classification of supernovae (SNe) based on optical photometric light curve information is essential in the upcoming era of wide-field time domain surveys, such as the Legacy Survey of Space and Time (LSST) conducted by the Rubin Observatory. Photometric classification can enable real-time identification of interesting events for extended multi-wavelength follow-up, as well as archival population studies. Here we present the complete sample of 5,243 "SN-like" light curves (in griz) from the Pan-STARRS1 Medium-Deep Survey (PS1-MDS). The PS1-MDS is similar to the planned LSST Wide-Fast-Deep survey in terms of cadence, filters and depth, making this a useful training set for the community. Using this dataset, we train a novel semi-supervised machine learning algorithm to photometrically classify 2,315 new SN-like light curves with host galaxy spectroscopic redshifts. Our algorithm consists of a random forest supervised classification step and a novel unsupervised step in which we introduce a recurrent autoencoder neural network (RAENN). Our final pipeline, dubbed SuperRAENN, has an accuracy of 87% across five SN classes (Type Ia, Ibc, II, IIn, SLSN-I). We find the highest accuracy rates for Type Ia SNe and SLSNe and the lowest for Type Ibc SNe. Our complete spectroscopically- and photometrically-classified samples break down into: 62.0% Type Ia (1839 objects), 19.8% Type II (553 objects), 4.8% Type IIn (136 objects), 11.7% Type Ibc (291 objects), and 1.6% Type I SLSNe (54 objects). Finally, we discuss how this algorithm can be modified for online LSST data streams., Comment: Submitted to ApJ; Companion paper to Hosseinzadeh et al.; Tables 1 and 2 available on ashleyvillar.com prior to publication

Social media corpora pose unique challenges and opportunities, including typically short document lengths and rich meta-data such as author characteristics and relationships. This creates great potential for systematic analysis of the enormous body of the users and thus provides implications for industrial strategies such as targeted marketing. Here we propose a novel and statistically principled method, clust-LDA, which incorporates authorship structure into the topical modeling, thus accomplishing the task of the topical inferences across documents on the basis of authorship and, simultaneously, the identification of groupings between authors. We develop an inference procedure for clust-LDA and demonstrate its performance on simulated data, showing that clust-LDA out-performs the "vanilla" LDA on the topic identification task where authors exhibit distinctive topical preference. We also showcase the empirical performance of clust-LDA based on a real-world social media dataset from Reddit.

This paper presents a large amount of observations for the $\delta$ Scuti star AN Lyncis carried out in 2001--2012. The extensive observations include two tri-continent campaigns coordinated in 2002 and 2011, respectively, and several single-site contributions throughout the period. The data in total have more than 104100 raw CCD frames and photoelectric records, which consist of 165 nights (about 816 hours) spanning over 3778 days. The final reduced light curves have more than 26500 data points (including those 3462 unpublished BYU data), from which we determined 306 new timings of maximum light. A time-dependent behaviour study based on all available data indicates cyclic amplitude variability as well as period change [for the main periodicity]. Orbital sinusoid fittings to $(O-C)$ residuals and pulsation amplitudes may account for their variations being caused by the light-time effect of AN Lyn in a binary system. The orbital period is about 26--30 years. Current results support the binarity of AN Lyn, first suspected by Zhou (2002). We further show the detailed time evolution structure of the pulsation of AN Lyn as function of both time and period through wavelet analyses., Comment: 23 figures, 5 tables, reduced data, maxima, and raw data available. Submitted initial manuscript, in revision in response to referee's report

Diffuse interstellar bands (DIBs) are absorption features commonly observed in optical/near-infrared spectra of stars and thought to be associated with polyatomic molecules that comprise a significant reservoir of organic material in the universe. However, the central wavelengths of almost all DIBs do not correspond with electronic transitions of known atomic or molecular species and the specific physical nature of their carriers remains inconclusive despite decades of observational, theoretical, and experimental research. It is well established that DIB carriers are located in the interstellar medium, but the recent discovery of time-varying DIBs in the spectra of the extragalactic supernova SN2012ap suggests that some may be created in massive star environments. Here we report evidence of short time-scale (10-60 d) changes in DIB absorption line substructure toward 3 of 17 massive stars observed as part of a pathfinder survey of variable DIBs conducted with the 1.5-m Tillinghast telescope and Tillinghast Reflector Echelle Spectrograph (TRES) at Fred L. Whipple Observatory. The detections are made in high-resolution optical spectra (R = 44000) having signal-to-noise ratios of 5-15 around the 5797 and 6614 angstrom features, and are considered significant but requiring further investigation. We find that these changes are potentially consistent with interactions between stellar winds and DIB carriers in close proximity. Our findings motivate a larger survey to further characterize these variations and may establish a powerful new method for probing the poorly understood physical characteristics of DIB carriers., Comment: 11 pages, 7 figures, accepted for publication in MNRAS

A primary goal of many undergraduate- and graduate-level courses in the physical sciences is to prepare students to engage in scientific research, or to prepare students for careers that leverage skillsets similar to those used by research scientists. Even for students who may not intend to pursue a career with these characteristics, exposure to the context of applications in modern research can be a valuable tool for teaching and learning. However, a persistent barrier to student participation in research is familiarity with the technical language, format, and context that academic researchers use to communicate research methods and findings with each other: the literature of the field. Astrobites, an online web resource authored by graduate students, has published brief and accessible summaries of more than 1,300 articles from the astrophysical literature since its founding in 2010. This article presents three methods for introducing students at all levels within the formal higher education setting to approaches and results from modern research. For each method, we provide a sample lesson plan that integrates content and principles from Astrobites, including step-by-step instructions for instructors, suggestions for adapting the lesson to different class levels across the undergraduate and graduate spectrum, sample student handouts, and a grading rubric., Comment: Accepted by AJP. 15 pages, 4 figures

Historically, light curve studies of supernovae (SNe) and other transient classes have focused on individual objects with copious and high signal-to-noise observations. In the nascent era of wide field transient searches, objects with detailed observations are decreasing as a fraction of the overall known SN population, and this strategy sacrifices the majority of the information contained in the data about the underlying population of transients. A population level modeling approach, simultaneously fitting all available observations of objects in a transient sub-class of interest, fully mines the data to infer the properties of the population and avoids certain systematic biases. We present a novel hierarchical Bayesian statistical model for population level modeling of transient light curves, and discuss its implementation using an efficient Hamiltonian Monte Carlo technique. As a test case, we apply this model to the Type IIP SN sample from the Pan-STARRS1 Medium Deep Survey, consisting of 18,837 photometric observations of 76 SNe, corresponding to a joint posterior distribution with 9,176 parameters under our model. Our hierarchical model fits provide improved constraints on light curve parameters relevant to the physical properties of their progenitor stars relative to modeling individual light curves alone. Moreover, we directly evaluate the probability for occurrence rates of unseen light curve characteristics from the model hyperparameters, addressing observational biases in survey methodology. We view this modeling framework as an unsupervised machine learning technique with the ability to maximize scientific returns from data to be collected by future wide field transient searches like LSST. \smallskip, Comment: Submitted; 10 pages, 11 figures, plus appendix with code

We present extensive radio and X-ray observations of SN\, 2012au, the energetic radio luminous supernova of type Ib that may be a link between subsets of hydrogen-poor superluminous and normal core-collapse supernovae. The observations closely follow models of synchrotron emission from shock heated circum-burst medium that has a wind density profile ($\rho \propto r^{-2}$). We infer a sub-relativistic velocity for the shock wave $v \approx 0.2\,c$ and a radius of $r \approx 1.4 \times 10^{16} \rm cm$ at 25 days after the estimated date of explosion. For a constant wind velocity of 1000 km/s we determine the constant mass loss rate of the progenitor to be $\dot{M} = 3.6 \times 10^{-6} \rm M_{\odot} yr^{-1}$, consistent with the estimates from X-ray observations. We estimate the total internal energy of the radio emitting material to be $E \approx 10^{47} \rm erg$, which is intermediate to SN\,1998bw and SN\,2002ap. Evolution of the radio light curves of SN\,2012au is consistent with interaction with a smoothly distributed circum-burst medium and absence of stellar shells ejected from previous outbursts out to $r \approx 10^{17} \rm cm$ from the supernova site. Based on this we conclude that the evolution of the SN\,2012au progenitor star was relatively quiet during the final years preceding explosion. We find that the bright radio emission from SN2012au was not dissimilar from other core collapse supernovae despite it's extraordinary optical properties. We speculate that it was the nature of the explosion that led to the unusual demise of the SN2012au progenitor star., Comment: 11 pages, 5 figures, submitted to ApJ

Educational technology has attained significant importance as a mechanism for supporting experiential learning of science concepts. However, the growth of this mechanism is limited by the significant time and technical expertise needed to develop such products, particularly in specialized fields of science. We sought to test whether interactive, educational, online software modules can be developed effectively by students as a curriculum component of an advanced science course. We discuss a set of fifteen such modules developed by Harvard University graduate students to demonstrate various concepts related to astronomy and physics. Their successful development of these modules demonstrates that online software tools for education and outreach on specialized topics can be produced while simultaneously fulfilling project-based learning objectives. We describe a set of technologies suitable for module development and present in detail four examples of modules developed by the students. We offer recommendations for incorporating educational software development within a graduate curriculum and conclude by discussing the relevance of this novel approach to new online learning environments like edX, Comment: 12 pages, 4 figures, 1 table; V2 as accepted by the Journal of Science Education and Technology; for more information visit http://ay201b.wordpress.com/topical-modules/2013-topical-modules/

There is a well known correlation between the mass and metallicity of star-forming galaxies. Because mass is correlated with luminosity, this relation is often exploited, when spectroscopy is not available, to estimate galaxy metallicities based on single band photometry. However, we show that galaxy color is typically more effective than luminosity as a predictor of metallicity. This is a consequence of the correlation between color and the galaxy mass-to-light ratio and the recently discovered correlation between star formation rate (SFR) and residuals from the mass-metallicity relation. Using Sloan Digital Sky Survey spectroscopy of 180,000 nearby galaxies, we derive "LZC relations," empirical relations between metallicity (in seven common strong line diagnostics), luminosity, and color (in ten filter pairs and four methods of photometry). We show that these relations allow photometric metallicity estimates, based on luminosity and a single optical color, that are 50% more precise than those made based on luminosity alone; galaxy metallicity can be estimated to within 0.05 - 0.1 dex of the spectroscopically-derived value depending on the metallicity diagnostic used. Including color information in metallicity estimates also reduces systematic biases for populations skewed toward high or low SFR environments, as we illustrate using the host galaxy of the supernova SN 2010ay. This new tool will lend more statistical power to studies of galaxy populations, such as supernova and gamma-ray burst (GRB) host environments, in ongoing and future wide field imaging surveys., Comment: 6 pages, 3 figures, 2 tables; additional plots and updated code available at http://www.cfa.harvard.edu/~nsanders/papers/LZC/summary.htm

The oxygen abundance gradients among nebular emission line regions in spiral galaxies have been used as important constraints for models of chemical evolution. We present the largest ever full-wavelength optical spectroscopic sample of emission line nebulae in a spiral galaxy (M31). We have collected spectra of 253 HII regions and 407 planetary nebulae with the Hectospec multi-fiber spectrograph of the MMT. We measure the line-of-sight extinction for 199 HII regions and 333 PNe; we derive oxygen abundance directly, based on the electron temperature, for 51 PNe; and we use strong line methods to estimate oxygen abundance for 192 HII regions and nitrogen abundance for 52 HII regions. The relatively shallow oxygen abundance gradient of the more extended HII regions in our sample is generally in agreement with the result of Zaritsky et al. (1994), based on only 19 M31 HII regions, but varies with the strong-line diagnostic employed. Our large sample size demonstrates that there is significant intrinsic scatter around this abundance gradient, as much as 3 times the systematic uncertainty in the strong line diagnostics. The intrinsic scatter is similar in the nitrogen abundances, although the gradient is significantly steeper. On small scales (deprojected distance < 0.5 kpc), HII regions exhibit local variations in oxygen abundance that are larger than 0.3 dex in 33% of neighboring pairs. We do not identify a significant oxygen abundance gradient among PNe, but we do find a significant gradient in the [N II] ratio that varies systematically with surface brightness. Our results underscore the complex and inhomogeneous nature of the ISM of M31, and our dataset illustrates systematic effects relevant to future studies of the metallicity gradients in nearby spiral galaxies., Comment: 22 pages, 14 figures, 5 tables. Accepted for publication in ApJ, full tables available at http://www.cfa.harvard.edu/~nsanders/papers/M31/summary.html

Because undergraduate participation in research is a longstanding and increasingly important aspect of the career path for future scientists, students can benefit from additional resources to introduce them to the culture and process of research. We suggest the adoption of the web resource Astrobites as a classroom tool to increase the preparation of undergraduate physics and astronomy students for careers in research. We describe the content and development of the website, discuss previous university courses that have made use of Astrobites, and suggest additional strategies for using Astrobites in the classroom., Comment: Published in the Astronomy Education Review

We present the largest spectroscopic study of the host environments of Type Ibc supernovae (SN Ibc) discovered exclusively by untargeted SN searches. Past studies of SN Ibc host environments have been biased towards high-mass, high-metallicity galaxies by focusing on SNe discovered in galaxy-targeted SN searches. Our new observations more than double the total number of spectroscopic stellar population age and metallicity measurements published for untargeted SN Ibc host environments, and extend to a median redshift about twice as large as previous statistical studies (z = 0.04). For the 12 SNe Ib and 21 SNe Ic in our metallicity sample, we find median metallicities of log(O/H)+12 = 8.48 and 8.61, respectively, but determine that the discrepancy in the full distribution of metallicities is not statistically significant. This median difference would correspond to only a small difference in the mass loss via metal-line driven winds (<30%), suggesting this does not play the dominant role in distinguishing SN Ib and Ic progenitors. However, the median metallicity of the 7 broad-lined SN Ic (SN Ic-BL) in our sample is significantly lower, log(O/H)+12 = 8.34. The age of the young stellar population of SN Ic-BL host environments also seems to be lower than for SN Ib and Ic, but our age sample is small. A synthesis of SN Ibc host environment spectroscopy to date does not reveal a significant difference in SN Ib and Ic metallicities, but reinforces the significance of the lower metallicities for SN Ic-BL. This combined sample demonstrates that galaxy-targeted SN searches introduce a significant bias for studies seeking to infer the metallicity distribution of SN progenitors, and we identify and discuss other systematic effects that play smaller roles. We discuss the path forward for making progress on SN Ibc progenitor studies in the LSST era., Comment: 27 pages, 12 Figures, V2 as accepted by ApJ, more information at http://www.cfa.harvard.edu/~nsanders/papers/Ibchosts/summary.html

We report on our serendipitous pre-discovery detection and detailed follow-up of the broad-lined Type Ic supernova (SN) 2010ay at z = 0.067 imaged by the Pan-STARRS1 3pi survey just ~4 days after explosion. The SN had a peak luminosity, M_R ~ -20.2 mag, significantly more luminous than known GRB-SNe and one of the most luminous SNe Ib/c ever discovered. The absorption velocity of SN 2010ay is v_Si ~ 19,000 km/s at ~40 days after explosion, 2-5 times higher than other broad-lined SNe and similar to the GRB-SN 2010bh at comparable epochs. Moreover, the velocity declines ~2 times slower than other SNe Ic-BL and GRB-SNe. Assuming that the optical emission is powered by radioactive decay, the peak magnitude implies the synthesis of an unusually large mass of 56 Ni, M_Ni = 0.9 M_solar. Modeling of the light-curve points to a total ejecta mass, M_ej ~ 4.7 M_sol, and total kinetic energy, E_K ~ 11x10^51 ergs. The ratio of M_Ni to M_ej is ~2 times as large for SN 2010ay as typical GRB-SNe and may suggest an additional energy reservoir. The metallicity (log(O/H)_PP04 + 12 = 8.19) of the explosion site within the host galaxy places SN 2010ay in the low-metallicity regime populated by GRB-SNe, and ~0.5(0.2) dex lower than that typically measured for the host environments of normal (broad-lined) Ic supernovae. We constrain any gamma-ray emission with E_gamma < 6x10^{48} erg (25-150 keV) and our deep radio follow-up observations with the Expanded Very Large Array rule out relativistic ejecta with energy, E > 10^48 erg. We therefore rule out the association of a relativistic outflow like those which accompanied SN 1998bw and traditional long-duration GRBs, but place less-stringent constraints on a weak afterglow like that seen from XRF 060218. These observations challenge the importance of progenitor metallicity for the production of a GRB, and suggest that other parameters also play a key role., Comment: 19 pages, 10 figures, V3 has revisions following referee's report; more information at http://www.cfa.harvard.edu/~nsanders/papers/2010ay/summary.html

To understand how best to use observations of Type Ia supernovae (SNe Ia) to obtain precise and accurate distances, we investigate the relations between spectra of SNe Ia and their intrinsic colors. Using a sample of 1630 optical spectra of 255 SNe, based primarily on data from the CfA Supernova Program, we examine how the velocity evolution and line strengths of Si II 6355 and Ca II H&K are related to the B-V color at peak brightness. We find that the maximum-light velocity of Si II 6355 and Ca II H&K and the maximum-light pseudo-equivalent width of Si II 6355 are correlated with intrinsic color, with intrinsic color having a linear relation with the Si II 6355 measurements. Ca II H&K does not have a linear relation with intrinsic color, but lower-velocity SNe tend to be intrinsically bluer. Combining the spectroscopic measurements does not improve intrinsic color inference. The intrinsic color scatter is larger for higher-velocity SNe Ia --- even after removing a linear trend with velocity --- indicating that lower-velocity SNe Ia are more "standard crayons." Employing information derived from SN Ia spectra has the potential to improve the measurements of extragalactic distances and the cosmological properties inferred from them., Comment: 23 pages, 21 figures, accepted by ApJ; data presented in the paper can be obtained at https://www.cfa.harvard.edu/~rfoley/data/