arXiv subjects

Tom Zick

Publications and source records attributed to Tom Zick.

At least 19 recordsLinked to original sources

arXiv2026

Underwriting the Agent Economy: The Blueprint for an AI Insurance Stack

From maritime trade to commercial nuclear power, insurance has been the enabler of major economic and technological developments by pricing risk, limiting downside, and spreading best practices. The emerging AI agent economy, projected to handle trillions of dollars in transactions by 2030, looks to be the next such development. Yet insurers' exposure to AI agent risk currently sits largely unpriced across existing insurance lines; between this silent coverage and growing exclusions, coverage is not fit for purpose. Furthermore, insurability is trending the wrong way: AI agent capabilities appear to be outpacing reliability, leading to rising incident severity; concentration among a few foundation model providers threatens correlated losses; and traditional actuarial modeling will struggle to keep pace with a technology evolving as rapidly as frontier AI. This report argues that affirmative AI coverage with limits in the billions is achievable by 2030, but only with industry-wide coordination. Drawing on successful historical precedents such as Underwriters Laboratories, the Closed Claims Project, and others, we lay out an eight-component AI insurance stack spanning incident data collection, catastrophe modeling, standards, contract design, risk selection, pricing, monitoring, and claims management. Building out this infrastructure is what will enable insurers to cover and manage AI agent risk sustainably and at scale. Finally, we discuss coverage for catastrophic risk from frontier AI ("AI CAT"), including CBRN, critical infrastructure collapse, and loss of control scenarios. Addressing these tail risks will require purpose-built instruments, potentially including a frontier model developer mutual, catastrophe bonds, bespoke liability regimes, and government backstops.

cs.CY↗

arXiv2025

Comparing Apples to Oranges: A Taxonomy for Navigating the Global Landscape of AI Regulation

AI governance has transitioned from soft law, such as national AI strategies and voluntary guidelines, to binding regulation at an unprecedented pace. This evolution has produced a complex legislative landscape: blurred definitions of "AI regulation" mislead the public and create a false sense of safety; divergent regulatory frameworks risk fragmenting international cooperation; and uneven access to key information heightens the danger of regulatory capture. Clarifying the scope and substance of AI regulation is vital to uphold democratic rights and align international AI efforts. We present a taxonomy to map the global landscape of AI regulation. Our framework targets essential metrics-technology or application-focused rules, horizontal or sectoral regulatory coverage, ex ante or ex post interventions, maturity of the digital legal landscape, enforcement mechanisms, and level of stakeholder participation-to classify the breadth and depth of AI regulation. We apply this framework to five early movers: the European Union's AI Act, the United States' Executive Order 14110, Canada's AI and Data Act, China's Interim Measures for Generative AI Services, and Brazil's AI Bill 2338/2023. We further offer an interactive visualization that distills these dense legal texts into accessible insights, highlighting both commonalities and differences. By delineating what qualifies as AI regulation and clarifying each jurisdiction's approach, our taxonomy reduces legal uncertainty, supports evidence-based policymaking, and lays the groundwork for more inclusive, globally coordinated AI governance.

cs.CY↗

arXiv2024

Personhood credentials: Artificial intelligence and the value of privacy-preserving tools to distinguish who is real online

Anonymity is an important principle online. However, malicious actors have long used misleading identities to conduct fraud, spread disinformation, and carry out other deceptive schemes. With the advent of increasingly capable AI, bad actors can amplify the potential scale and effectiveness of their operations, intensifying the challenge of balancing anonymity and trustworthiness online. In this paper, we analyze the value of a new tool to address this challenge: "personhood credentials" (PHCs), digital credentials that empower users to demonstrate that they are real people -- not AIs -- to online services, without disclosing any personal information. Such credentials can be issued by a range of trusted institutions -- governments or otherwise. A PHC system, according to our definition, could be local or global, and does not need to be biometrics-based. Two trends in AI contribute to the urgency of the challenge: AI's increasing indistinguishability from people online (i.e., lifelike content and avatars, agentic activity), and AI's increasing scalability (i.e., cost-effectiveness, accessibility). Drawing on a long history of research into anonymous credentials and "proof-of-personhood" systems, personhood credentials give people a way to signal their trustworthiness on online platforms, and offer service providers new tools for reducing misuse by bad actors. In contrast, existing countermeasures to automated deception -- such as CAPTCHAs -- are inadequate against sophisticated AI, while stringent identity verification solutions are insufficiently private for many use-cases. After surveying the benefits of personhood credentials, we also examine deployment risks and design challenges. We conclude with actionable next steps for policymakers, technologists, and standards bodies to consider in consultation with the public.

cs.CY↗

arXiv2024

AI Procurement Checklists: Revisiting Implementation in the Age of AI Governance

Public sector use of AI has been quietly on the rise for the past decade, but only recently have efforts to regulate it entered the cultural zeitgeist. While simple to articulate, promoting ethical and effective roll outs of AI systems in government is a notoriously elusive task. On the one hand there are hard-to-address pitfalls associated with AI-based tools, including concerns about bias towards marginalized communities, safety, and gameability. On the other, there is pressure not to make it too difficult to adopt AI, especially in the public sector which typically has fewer resources than the private sector$\unicode{x2014}$conserving scarce government resources is often the draw of using AI-based tools in the first place. These tensions create a real risk that procedures built to ensure marginalized groups are not hurt by government use of AI will, in practice, be performative and ineffective. To inform the latest wave of regulatory efforts in the United States, we look to jurisdictions with mature regulations around government AI use. We report on lessons learned by officials in Brazil, Singapore and Canada, who have collectively implemented risk categories, disclosure requirements and assessments into the way they procure AI tools. In particular, we investigate two implemented checklists: the Canadian Directive on Automated Decision-Making (CDADM) and the World Economic Forum's AI Procurement in a Box (WEF). We detail three key pitfalls around expertise, risk frameworks and transparency, that can decrease the efficacy of regulations aimed at government AI use and suggest avenues for improvement.

cs.CY↗

arXiv2024

RewardBench: Evaluating Reward Models for Language Modeling

Reward models (RMs) are at the crux of successfully using RLHF to align pretrained models to human preferences, yet there has been relatively little study that focuses on evaluation of those models. Evaluating reward models presents an opportunity to understand the opaque technologies used for alignment of language models and which values are embedded in them. Resources for reward model training and understanding are sparse in the nascent open-source community around them. To enhance scientific understanding of reward models, we present RewardBench, a benchmark dataset and code-base for evaluation. The RewardBench dataset is a collection of prompt-chosen-rejected trios spanning chat, reasoning, and safety, to benchmark how reward models perform on challenging, structured and out-of-distribution queries. We create specific comparison datasets for RMs that have subtle, but verifiable reasons (e.g. bugs, incorrect facts) why one answer should be preferred to another. On the RewardBench leaderboard, we evaluate reward models trained with a variety of methods, such as the direct MLE training of classifiers and the implicit reward modeling of Direct Preference Optimization (DPO). We present many findings on propensity for refusals, reasoning limitations, and instruction following shortcomings of various reward models towards a better understanding of the RLHF process.

cs.LG↗

arXiv2023

The History and Risks of Reinforcement Learning and Human Feedback

Reinforcement learning from human feedback (RLHF) has emerged as a powerful technique to make large language models (LLMs) easier to use and more effective. A core piece of the RLHF process is the training and utilization of a model of human preferences that acts as a reward function for optimization. This approach, which operates at the intersection of many stakeholders and academic disciplines, remains poorly understood. RLHF reward models are often cited as being central to achieving performance, yet very few descriptors of capabilities, evaluations, training methods, or open-source models exist. Given this lack of information, further study and transparency is needed for learned RLHF reward models. In this paper, we illustrate the complex history of optimizing preferences, and articulate lines of inquiry to understand the sociotechnical context of reward models. In particular, we highlight the ontological differences between costs, rewards, and preferences at stake in RLHF's foundations, related methodological tensions, and possible research directions to improve general understanding of how reward models function.

cs.CY↗

arXiv2022

The MOSDEF Survey: Towards a Complete Census of the z ~ 2.3 Star-forming Galaxy Population

We analyze the completeness of the MOSDEF survey, in which z ~ 2 galaxies were selected for rest-optical spectroscopy from well-studied HST extragalactic legacy fields down to a fixed rest-optical magnitude limit (H_AB = 24.5). The subset of z ~ 2 MOSDEF galaxies with high signal-to-noise (S/N) emission-line detections analyzed in previous work represents a small minority (<10%) of possible z ~ 2 MOSDEF targets. It is therefore crucial to understand how representative this high S/N subsample is, while also more fully exploiting the MOSDEF spectroscopic sample. Using spectral-energy-distribution (SED) models and rest-optical spectral stacking, we compare the MOSDEF z ~ 2 high S/N subsample with the full MOSDEF sample of z ~ 2 star-forming galaxies with redshifts, the latter representing an increase in sample size of more than a factor of three. We find that both samples have similar emission-line properties, in particular in terms of the magnitude of the offset from the local star-forming sequence on the [N II] BPT diagram. There are small differences in median host galaxy properties, including the stellar mass (M_*), star-formation rate (SFR) and specific SFR (sSFR), and UVJ colors; however, these offsets are minor considering the wide spread of the distributions. Using SED modeling, we also demonstrate that the sample of z ~ 2 star-forming galaxies observed by the MOSDEF survey is representative of the parent catalog of available such targets. We conclude that previous MOSDEF results on the evolution of star-forming galaxy emission-line properties were unbiased relative to the parent z ~ 2 galaxy population.

astro-ph.GA↗

arXiv2022

The MOSDEF Survey: A New View of a Remarkable z=1.89 Merger

We present a detailed study of a galaxy merger taking place at $z=1.89$ in the GOODS-S field. Here we analyze Keck/MOSFIRE spectroscopic observations from the MOSFIRE Deep Evolution Field (MOSDEF) survey along with multi-wavelength photometry assembled by the 3D-HST survey. The combined dataset is modeled to infer the past star-formation histories (SFHs) of both merging galaxies. They are found to be massive, with log$_{10}(M_{\ast}/M_{\odot}) > 11$, with a close mass ratio satisfying the typical major-merger definition. Additionally, in the context of delayed-$\tau$ models, GOODS-S 43114 and GOODS-S 43683 have similar SFHs and low star-formation rates (log$_{10}$(SFR(SED)/$M_{\odot}/\rm{yr}^{-1}$) $<$ 1.0) compared to their past averages. The best-fit model SEDs show elevated H$\delta_{\rm{A}}$ values for both galaxies, indicating that their stellar spectra are dominated by A-type stars, and that star formation peaked $\sim0.5-1$ Gyr ago and has recently declined. Additionally, based on SED fitting both merging galaxies turned on and shut off star formation within a few hundred Myr of each other, suggesting that their bursts of star formation may be linked. Combining the SFHs and H$\delta_{\rm{A}}$ results with recent galaxy merger simulations, we infer that these galaxies have recently completed their first pericentric passage and are moving apart. Finally, the relatively low second velocity moment of GOODS-S 43114 given its stellar mass, suggests a disk-like structure. However, including the geometry of the galaxy in the modeling does not completely resolve the discrepancy between the dynamical and stellar masses. Future work is needed to resolve this inconsistency in mass.

astro-ph.GA↗

arXiv2022

Reward Reports for Reinforcement Learning

Building systems that are good for society in the face of complex societal effects requires a dynamic approach. Recent approaches to machine learning (ML) documentation have demonstrated the promise of discursive frameworks for deliberation about these complexities. However, these developments have been grounded in a static ML paradigm, leaving the role of feedback and post-deployment performance unexamined. Meanwhile, recent work in reinforcement learning has shown that the effects of feedback and optimization objectives on system behavior can be wide-ranging and unpredictable. In this paper we sketch a framework for documenting deployed and iteratively updated learning systems, which we call Reward Reports. Taking inspiration from various contributions to the technical literature on reinforcement learning, we outline Reward Reports as living documents that track updates to design choices and assumptions behind what a particular automated system is optimizing for. They are intended to track dynamic phenomena arising from system deployment, rather than merely static properties of models or data. After presenting the elements of a Reward Report, we discuss a concrete example: Meta's BlenderBot 3 chatbot. Several others for game-playing (DeepMind's MuZero), content recommendation (MovieLens), and traffic control (Project Flow) are included in the appendix.

cs.LG↗

arXiv2022

Choices, Risks, and Reward Reports: Charting Public Policy for Reinforcement Learning Systems

In the long term, reinforcement learning (RL) is considered by many AI theorists to be the most promising path to artificial general intelligence. This places RL practitioners in a position to design systems that have never existed before and lack prior documentation in law and policy. Public agencies could intervene on complex dynamics that were previously too opaque to deliberate about, and long-held policy ambitions would finally be made tractable. In this whitepaper we illustrate this potential and how it might be technically enacted in the domains of energy infrastructure, social media recommender systems, and transportation. Alongside these unprecedented interventions come new forms of risk that exacerbate the harms already generated by standard machine learning tools. We correspondingly present a new typology of risks arising from RL design choices, falling under four categories: scoping the horizon, defining rewards, pruning information, and training multiple agents. Rather than allowing RL systems to unilaterally reshape human domains, policymakers need new mechanisms for the rule of reason, foreseeability, and interoperability that match the risks these systems pose. We argue that criteria for these choices may be drawn from emerging subfields within antitrust, tort, and administrative law. It will then be possible for courts, federal and state agencies, and non-governmental organizations to play more active roles in RL specification and evaluation. Building on the "model cards" and "datasheets" frameworks proposed by Mitchell et al. and Gebru et al., we argue the need for Reward Reports for AI systems. Reward Reports are living documents for proposed RL deployments that demarcate design choices.

cs.LG↗

arXiv2021

Reconciling the Results of the z~2 MOSDEF and KBSS-MOSFIRE Surveys

The combination of the MOSDEF and KBSS-MOSFIRE surveys represents the largest joint investment of Keck/MOSFIRE time to date, with ~3000 galaxies at 1.4<=z<=3.8, roughly half of which are at z~2. MOSDEF is photometric- and spectroscopic-redshift selected with a rest-optical magnitude limit, while KBSS-MOSFIRE is primarily selected based on rest-UV colors and a rest-UV magnitude limit. Analyzing both surveys in a uniform manner with consistent spectral-energy-distribution (SED) models, we find that the MOSDEF z~2 targeted sample has a higher median M_* and redder rest U-V color than the KBSS-MOSFIRE z~2 targeted sample, and a smaller median SED-based SFR and sSFR (SFR(SED) and sSFR(SED)). Specifically, MOSDEF targeted a larger population of red galaxies with U-V and V-J >=1.25, while KBSS-MOSFIRE contains more young galaxies with intense star formation. Despite these differences in the z~2 targeted samples, the subsets of the surveys with multiple emission lines detected and analyzed in previously published work are much more similar. All median host-galaxy properties with the exception of stellar population age -- i.e., M_*, SFR(SED), sSFR(SED), A_V, and UVJ colors -- agree within the uncertainties. Additionally, when uniform emission-line fitting and stellar Balmer absorption correction techniques are applied, there is no significant offset between the two samples in the [OIII]$\lambda$5008/H$\beta$ vs. [NII]$\lambda$6585/H$\alpha$ diagnostic diagram, in contrast to previously-reported discrepancies. We can now combine the MOSDEF and KBSS-MOSFIRE surveys to form the largest z~2 sample with moderate-resolution rest-optical spectra and construct the fundamental scaling relations of star-forming galaxies during this important epoch.

astro-ph.GA↗

arXiv2021

Axes for Sociotechnical Inquiry in AI Research

The development of artificial intelligence (AI) technologies has far exceeded the investigation of their relationship with society. Sociotechnical inquiry is needed to mitigate the harms of new technologies whose potential impacts remain poorly understood. To date, subfields of AI research develop primarily individual views on their relationship with sociotechnics, while tools for external investigation, comparison, and cross-pollination are lacking. In this paper, we propose four directions for inquiry into new and evolving areas of technological development: value--what progress and direction does a field promote, optimization--how the defined system within a problem formulation relates to broader dynamics, consensus--how agreement is achieved and who is included in building it, and failure--what methods are pursued when the problem specification is found wanting. The paper provides a lexicon for sociotechnical inquiry and illustrates it through the example of consumer drone technology.

cs.CY↗

arXiv2021

The MOSDEF Survey: The Mass-Metallicity relationship and the existence of the FMR at z~1.5

We analyze the rest-optical emission-line ratios of z~1.5 galaxies drawn from the MOSFIRE Deep Evolution Field (MOSDEF) survey. Using composite spectra we investigate the mass-metallicity relation (MZR) at z~1.5 and measure its evolution to z=0. When using gas-phase metallicities based on the N2 line ratio, we find that the MZR evolution from z~1.5 to z=0 depends on stellar mass, evolving by $\Delta\rm log(\rm O/H)\sim0.25$ dex at $M_*<10^{9.75}M_{\odot}$ down to $\Delta\rm log(\rm O/H)\sim0.05$ at $M_*>10^{10.5}M_{\odot}$. In contrast, the O3N2-based MZR shows a constant offset of $\Delta\rm log(\rm O/H)\sim0.30$ across all masses, consistent with previous MOSDEF results based on independent metallicity indicators, and suggesting that O3N2 provides a more robust metallicity calibration for our z~1.5 sample. We investigated the secondary dependence of the MZR on SFR by measuring correlated scatter about the mean $M_*$-specific SFR and $M_*-\log(\rm O3N2)$ relations. We find an anti-correlation between $\log(\rm O/H)$ and sSFR offsets, indicating the presence of a $M_*$-SFR-Z relation, though with limited significance. Additionally, we find that our z~1.5 stacks lie along the z=0 metallicity sequence at fixed $\mu=\log(M_*/M_{\odot})-0.6\times\log(\rm SFR / M_{\odot} yr^{-1})$ suggesting that the z~1.5 stacks can be described by the z=0 fundamental metallicity relation (FMR). However, using different calibrations can shift the calculated metallicities off of the local FMR, indicating that appropriate calibrations are essential for understanding metallicity evolution with redshift. Finally, understanding how [NII]/H$\alpha$ scales with galaxy properties is crucial to accurately describe the effects of blended [NII] and H$\alpha$ on redshift and H$\alpha$ flux measurements in future large surveys utilizing low-resolution spectra such as with Euclid and the Roman Space Telescope.

astro-ph.GA↗

arXiv2021

AI Development for the Public Interest: From Abstraction Traps to Sociotechnical Risks

Despite interest in communicating ethical problems and social contexts within the undergraduate curriculum to advance Public Interest Technology (PIT) goals, interventions at the graduate level remain largely unexplored. This may be due to the conflicting ways through which distinct Artificial Intelligence (AI) research tracks conceive of their interface with social contexts. In this paper we track the historical emergence of sociotechnical inquiry in three distinct subfields of AI research: AI Safety, Fair Machine Learning (Fair ML) and Human-in-the-Loop (HIL) Autonomy. We show that for each subfield, perceptions of PIT stem from the particular dangers faced by past integration of technical systems within a normative social order. We further interrogate how these histories dictate the response of each subfield to conceptual traps, as defined in the Science and Technology Studies literature. Finally, through a comparative analysis of these currently siloed fields, we present a roadmap for a unified approach to sociotechnical graduate pedagogy in AI.

cs.CY↗

arXiv2021

The MOSDEF Survey: Environmental dependence of the gas-phase metallicity of galaxies at $1.4 \leq z \leq 2.6$

Using the near-IR spectroscopy of the MOSFIRE Deep Evolution Field (MOSDEF) survey, we investigate the role of local environment in the gas-phase metallicity of galaxies. The local environment measurements are derived from accurate and uniformly calculated photometric redshifts with well-calibrated probability distributions. Based on rest-frame optical emission lines, [NII]$\lambda6584$ and H$\alpha$, we measure gas-phase oxygen abundance of 167 galaxies at $1.37\leq z\leq1.7$ and 303 galaxies at $2.09\leq z\leq2.61$, located in diverse environments. We find that at $z\sim1.5$, the average metallicity of galaxies in overdensities with $M_*\sim10^{9.8}M_\odot, 10^{10.2}M_\odot$ and $10^{10.8}M_\odot$ is higher relative to their field counterparts by $0.094\pm0.051$, $0.068\pm0.028$ and $0.052\pm0.043$ dex, respectively. However, this metallicity enhancement does not exist at higher redshift, $z\sim2.3$, where, compared to the field galaxies, we find $0.056\pm0.043$, $0.056\pm0.028$ and $0.096\pm 0.034$ dex lower metallicity for galaxies in overdense environments with $M_*\sim10^{9.8}M_\odot, 10^{10.2}M_\odot$ and $10^{10.7}M_\odot$, respectively. Our results suggest that, at $1.37\leq z\leq2.61$, the variation of mass-metallicity relation with local environment is small ($<0.1$dex), and reverses at $z\sim2$. Our results support the hypothesis that, at the early stages of cluster formation, owing to efficient gas cooling, galaxies residing in overdensities host a higher fraction of pristine gas with prominent primordial gas accretion, which lowers their gas-phase metallicity compared to their coeval field galaxies. However, as the Universe evolves to lower redshifts ($z\lesssim2$), shock-heated gas in overdensities cannot cool down efficiently, and galaxies become metal-rich rapidly due to the suppression of pristine gas inflow and re-accretion of metal-enriched outflows in overdensities.

astro-ph.GA↗

arXiv2020

The MOSDEF Survey: The First Direct Measurements of the Nebular Dust Attenuation Curve at High Redshift

We use a sample of 532 star-forming galaxies at redshifts $z\sim 1.4-2.6$ with deep rest-frame optical spectra from the MOSFIRE Deep Evolution Field (MOSDEF) survey to place the first constraints on the nebular attenuation curve at high redshift. Based on the first five low-order Balmer emission lines detected in the composite spectra of these galaxies (${\rm H\alpha}$ through ${\rm H\epsilon}$), we derive a nebular attenuation curve that is similar in shape to that of the Galactic extinction curve, suggesting that the dust covering fraction and absorption/scattering properties along the lines-of-sight to massive stars at high redshift are similar to those of the average Milky Way sightline. The curve derived here implies nebular reddening values that are on average systematically larger than those derived for the stellar continuum. In the context of stellar population synthesis models that include the effects of stellar multiplicity, the difference in reddening of the nebular lines and stellar continuum may imply molecular cloud crossing timescales that are a factor of $\gtrsim 3\times$ longer than those inferred for local molecular clouds, star-formation rates that are constant or increasing with time such that newly-formed and dustier OB associations always dominate the ionizing flux, and/or that the dust responsible for reddening the nebular emission may be associated with non-molecular (i.e., ionized and neutral) phases of the ISM. Our analysis points to a variety of investigations of the nebular attenuation curve that will be enabled with the next generation of ground- and space-based facilities.

astro-ph.GA↗

arXiv2020

The MOSDEF Survey: The Evolution of the Mass-Metallicity Relation from $z=0$ to $z\sim3.3$

We investigate the evolution of galaxy gas-phase metallicity (O/H) over the range $z=0-3.3$ using samples of $\sim300$ galaxies at $z\sim2.3$ and $\sim150$ galaxies at $z\sim3.3$ from the MOSDEF survey. This analysis crucially utilizes different metallicity calibrations at $z\sim0$ and $z>1$ to account for evolving ISM conditions. We find significant correlations between O/H and stellar mass ($M_*$) at $z\sim2.3$ and $z\sim3.3$. The low-mass power law slope of the mass-metallicity relation is remarkably invariant over $z=0-3.3$, such that $\textrm{O/H}\propto M_*^{0.30}$ at all redshifts in this range. At fixed $M_*$, O/H decreases with increasing redshift as dlog(O/H)/d$z=-0.11\pm0.02$. We find no evidence that the fundamental metallicity relation between $M_*$, O/H, and star-formation rate (SFR) evolves out to $z\sim3.3$, with galaxies at $z\sim2.3-3.3$ having O/H within 0.04~dex of local galaxies matched in $M_*$ and SFR on average. We employ analytic chemical evolution models to place constraints on the mass and metal loading factors of galactic outflows. The efficiency of metal removal increases toward lower $M_*$ at fixed redshift, and toward higher redshift at fixed $M_*$. These models suggest that the slope of the mass-metallicity relation is set by the scaling of the metal loading factor of outflows with $M_*$, not by the change in gas fraction as a function of $M_*$. The evolution toward lower O/H at fixed $M_*$ with increasing redshift is driven by both higher gas fraction (leading to stronger dilution of ISM metals) and higher metal removal efficiency, with models suggesting that both effects contribute approximately equally to the observed evolution. These results suggest that the processes governing the smooth baryonic growth of galaxies via gas flows and star formation hold in the same form over at least the past 12~Gyr.

astro-ph.GA↗

arXiv2020

The MOSDEF Survey: An Improved Voronoi Binning Technique on Spatially Resolved Stellar Populations at z~2

We use a sample of 350 star-forming galaxies at $1.25<z<2.66$ from the MOSFIRE Deep Evolution Field survey to demonstrate an improved Voronoi binning technique that we use to study the properties of resolved stellar populations in $z\sim2$ galaxies. Stellar population and dust maps are constructed from the high-resolution CANDELS/3D-HST multi-band imaging. Rather than constructing the layout of resolved elements (i.e., Voronoi bins) from the S/N distribution of the $H_{160}$-band alone, we introduce a modified Voronoi binning method that additionally incorporates the S/N distribution of several resolved filters. The SED-derived resolved E(B-V)$_{\text{stars}}$, stellar population ages, SFRs, and stellar masses that are inferred from the Voronoi bins constructed from multiple filters are generally consistent with the properties inferred from the integrated photometry within the uncertainties, with the exception of the inferred E(B-V)$_{\text{stars}}$ from our $z\sim1.5$ sample due to their UV slopes being unconstrained by the resolved photometry. The results from our multi-filter Voronoi binning technique are compared to those derived from a "traditional" single-filter Voronoi binning approach. We find that single-filter binning produces inferred E(B-V)$_{\text{stars}}$ that are systematically redder by 0.02 mag on average, but could differ by up to 0.20 mag, and could be attributed to poorly constrained resolved photometry covering the UV slope. Overall, we advocate that our methodology produces more reliable SED-derived parameters due to the best-fit resolved SEDs being better constrained at all resolved wavelengths--particularly those covering the UV slope.

astro-ph.GA↗