Search

Extant work shows that generative AI models such as GPT-3.5 and 4 perpetuate social stereotypes and biases. One concerning but less explored source of bias is ideology. Do GPT models take ideological stances on politically sensitive topics? In this article, we provide an original approach to identifying ideological bias in generative models, showing that bias can stem from both the training data and the filtering algorithm. We leverage linguistic variation in countries with contrasting political attitudes to evaluate bias in average GPT responses to sensitive political topics in those languages. First, we find that GPT output is more conservative in languages that map well onto conservative societies (i.e., Polish), and more liberal in languages used uniquely in liberal societies (i.e., Swedish). This result provides strong evidence of training data bias in GPT models. Second, differences across languages observed in GPT-3.5 persist in GPT-4, even though GPT-4 is significantly more liberal due to OpenAI's filtering policy. Our main takeaway is that generative model training must focus on high-quality, curated datasets to reduce bias, even if it entails a compromise in training data size. Filtering responses after training only introduces new biases and does not remove the underlying training biases.

This article presents the Political Deepfakes Incidents Database (PDID), a collection of politically-salient deepfakes, encompassing synthetically-created videos, images, and less-sophisticated `cheapfakes.' The project is driven by the rise of generative AI in politics, ongoing policy efforts to address harms, and the need to connect AI incidents and political communication research. The database contains political deepfake content, metadata, and researcher-coded descriptors drawn from political science, public policy, communication, and misinformation studies. It aims to help reveal the prevalence, trends, and impact of political deepfakes, such as those featuring major political figures or events. The PDID can benefit policymakers, researchers, journalists, fact-checkers, and the public by providing insights into deepfake usage, aiding in regulation, enabling in-depth analyses, supporting fact-checking and trust-building efforts, and raising awareness of political deepfakes. It is suitable for research and application on media effects, political discourse, AI ethics, technology governance, media literacy, and countermeasures., Comment: 6 pages, 1 table

Quality child care enables parents to work or go to school while also providing young children with the early childhood education experiences needed for healthy development. The Child Care and Development Block Grant (CCDBG) is the primary federal program that provides funding for child care assistance for low-income working parents. Child care assistance is a vital public investment that increases the sustainability of employment for low-income parents and provides stability for parents struggling to gain economic security. It also allows many parents to access higher quality care than they could otherwise afford. In spite of several years of post-recession economic recovery, the percentage of Americans living in poverty remains high. The majority of parents with young children work. More than 30 percent of poor children and half of low-income children--living in families earning less than 200 percent of the federal poverty level--lived in families with at least one worker employed full-time, year round. For parents working in low-wage jobs, paying the high costs of child care is a struggle. The average annual costs of center-based care for a 4-year-old ranges from $4,312 in Mississippi to $12,355 in New York. Nationwide, families living below poverty who pay for child care spend approximately 30 percent of their income, which is significantly higher than families not in poverty who pay 8 percent of their incomes. This brief highlights why child care assistance is an essential work support program for low-income working families and provides evidence that these subsidies are associated with sustainable employment for parents and improved child outcomes. It makes the case for why additional investments at the federal and state level are critical.

Since 1965, the Head Start program has served low-income 3- and 4-year-old children and their families with comprehensive early education and support services. Programs provide services focused on the "whole child," including early education addressing cognitive, developmental, and socio-emotional needs; medical and dental screenings and referrals; nutritional services; parental involvement activities; referrals to social service providers for the entire family; and mental health services. All Head Start programs are required to complete the Program Information Report (PIR) on an annual basis. The PIR collects data on all children and pregnant women who participate in Head Start at any point during the program year, including those who do not complete the year. This fact sheet uses information reported through PIR to describe the children and families enrolled in the Head Start preschool program and the services provided to them during the 2013-2014 program year.

In 1994, the federal Early Head Start (EHS) program was created to address the comprehensive needs of low-income pregnant women and children under age 3. EHS was launched almost 30 years after Head Start was established in 1965 to serve low-income 3- and 4-year-old children and their families with comprehensive early education and support services. Both programs provide services focused on the "whole child," including early education addressing cognitive, developmental, and socio-emotional needs; medical and dental screenings and referrals; nutritional services; parental involvement activities; referrals to social service providers for the entire family; and mental health services. All Head Start programs are required to complete the Program Information Report (PIR) on an annual basis. This fact sheet uses information reported through the PIR to describe the children and families served in Early Head Start and the services provided to them during the 2013-2014 program year.

We have carried out a Spitzer survey for brown dwarf (BD) disks in the ~5 Myr old Upper Scorpius (UpSco) star forming region, using IRS spectroscopy from 8 to 12\mu m and MIPS photometry at 24\mu m. Our sample consists of 35 confirmed very low mass members of UpSco. Thirteen objects in this sample show clear excess flux at 24\mu m, explained by dust emission from a circum-sub-stellar disk. Objects without excess emission either have no disks at all or disks with inner opacity holes of at least ~5 AU radii. Our disk frequency of 37\pm 9% is higher than what has been derived previously for K0-M5 stars in the same region (on a 1.8 sigma confidence level), suggesting a mass-dependent disk lifetime in UpSco. The clear distinction between objects with and without disks as well as the lack of transition objects shows that disk dissipation inside 5 AU occurs rapidly, probably on timescales of <~10^5 years. For the objects with disks, most SEDs are uniformly flat with flux levels of a few mJy, well modeled as emission from dusty disks affected by dust settling to the midplane, which also provides indirect evidence for grain growth. The silicate feature around 10\mu m is either absent or weak in our SEDs, arguing for a lack of hot, small dust grains. Compared with younger objects in Taurus, BD disks in UpSco show less flaring. Taken together, these results clearly demonstrate that we see disks in an advanced evolutionary state: Dust settling and grain growth are ubiquituous in circum-sub-stellar disks at ages of 5 Myr, arguing for planet forming processes in BD disks. For almost all our targets, results from high-resolution spectroscopy and high-spatial resolution imaging have been published before, thus providing a large sample of BDs for which information about disks, accretion, and binarity is available. (abridged), Comment: 39 pages, 7 figures, accepted for publication in ApJ

Head Start programs provide poor children and their families with comprehensive early education and support services. Each year, programs are required to submit a Program Information Report (PIR) to the Office of Head Start on participating children, pregnant women, and families, as well as the staff serving the Head Start population. In 2013, the Head Start program served more than 1.1 million children, approximately 166,000 of whom were under the age of 3, and 15,400 pregnant women through Head Start (HS) preschool, Early Head Start (EHS), and Migrant and Seasonal Head Start (MSHS) programs combined. HS preschool served 85 percent of all federally funded slots, with EHS and MSHS each serving 12 and 3 percent of slots respectively. The 2013 appropriation for all Head Start programs was just over $8 billion; however, federal budget cuts (known as sequestration) were in effect during this school year, resulting in reduced funding for this and other federal programs. While some Head Start programs managed sequestration by reducing the number of children served, others cut back schedules or made other cuts in their budgets.

In 1994, the federal Early Head Start (EHS) program was created to address the comprehensive needs of low-income pregnant women and children under age 3. EHS was launched almost 30 years after Head Start was established in 1965 to serve low-income 3- and 4-year-old children and their families with comprehensive early education and support services. Both programs provide services focused on the "whole child," including early education addressing cognitive, developmental, and socio-emotional needs; medical and dental screenings and referrals; nutritional services; parental involvement activities; referrals to social service providers for the entire family; and mental health services. All Head Start programs are required to complete the Program Information Report (PIR) on an annual basis. This fact sheet uses information reported through the PIR to describe the children and families served in Early Head Start and the services provided to them during the 2012-2013 program year. EHS spending in 2013 totaled $1.3 billion. In 2013, the EHS program served 150,100 children under age 3 and 15,459 pregnant women through 1,020 grantees/delegates throughout the country. Additional key findings from the 2013 PIR are provided herein.

We present Monte Carlo scattered light models of a warped disk that reproduce the observed photopolarimetric variability of the classical T Tauri star AA Tau. For a system inclination of 75 deg and using an analytic description for a warped inner disk, we find that the shape and amplitude of the photopolarimetric variability are reproduced with a warp which occults the star, located at 0.07 au, amplitude 0.016 au, extending over radial and azimuthal ranges 0.0084 au and 145 deg. We also show a time sequence of high spatial resolution scattered light images, showing a dark shadow cast by the warp sweeping round the disk. Using a modified smooth particle hydrodynamics code, we find that a stellar dipole magnetic field of strength 5.2 kG, inclined at 30 deg to the stellar rotation axis can reproduce the required disk warping to explain AA Tau's photopolarimetric variability., Comment: 9 Pages, 10 Figures. Accepted for publishing in MNRAS

We present synthetic spectra for circumstellar disks that are heated by radiation from a central brown dwarf. Under the assumption of vertical hydrostatic equilibrium, our models yield scaleheights for brown dwarf disks in excess of three times those derived for classical T Tauri (CTTS) disks. If the near-IR excess emission observed from brown dwarfs is indeed due to circumstellar disks, then the large scaleheights we find could have a significant impact on the optical and near-IR detectability of such systems. Our radiation transfer calculations show that such highly flared disks around brown dwarfs will result in a large fraction of obscured sources due to extinction of direct starlight by the disk over a wide range of sightlines. The obscured fraction for a 'typical' CTTS is less than 20%. We show that the obscured fraction for brown dwarfs may be double that for CTTS, but this depends on stellar and disk mass. We also comment on possible confusion in identifying brown dwarfs via color-magnitude diagrams: edge-on CTTS display similar colors and magnitudes as a face-on brown dwarf plus disk systems., Comment: 13 pages, 13 figures, accepted for publication in MNRAS

We present images of the young stellar object ASR 41 in the NGC 1333 star forming region at the wavelengths of H_alpha and [SII] and in the I, J, H, and K-bands. ASR 41 has the near-infrared morphology of an edge-on disk object, but appears an order of magnitude larger than typical systems of this kind. We also present detailed models of the scattering and radiative transfer in systems consisting of a young star surrounded by a proto-planetary disk, and the whole system being embedded in either an infalling envelope or a uniform molecular cloud. The best fit to the observed morphology can be achieved with a disk of approx. 200 AU diameter, immersed in a low density cloud. The low cloud density is necessary to stay below the sub-mm flux upper limits and to preserve the shadow cast by the disk via single scattering. The results demonstrate that ASR 41 is probably not inherently different from typical edge-on disk objects, and that its large apparent size is due to the shadow of a much smaller disk being projected into the surrounding dusty molecular material.

Migrant and Seasonal Head Start (MSHS) provides child care and other services to migrants to ensure that young children from birth to age 5 are not with their parents in the fields, where they can be exposed to pesticides, hazardous equipment, extreme heat, and other health dangers. MSHS provides services focused on the "whole child," including early education addressing cognitive, developmental, and socio-emotional needs; medical and dental screenings and referrals; nutritional services; parental involvement activities; referrals to social service providers for the entire family; and mental health services. MSHS programs serve children from birth to five years old. All Head Start programs are required to complete the Program Information Report (PIR) on an annual basis. This fact sheet uses information reported through the PIR to describe the children and families served in Migrant and Seasonal Head Start and the services provided to them during the 2013-2014 program year.