Search

Recommender systems often rely on models which are trained to maximize accuracy in predicting user preferences. When the systems are deployed, these models determine the availability of content and information to different users. The gap between these objectives gives rise to a potential for unintended consequences, contributing to phenomena such as filter bubbles and polarization. In this work, we consider directly the information availability problem through the lens of user recourse. Using ideas of reachability, we propose a computationally efficient audit for top-$N$ linear recommender models. Furthermore, we describe the relationship between model complexity and the effort necessary for users to exert control over their recommendations. We use this insight to provide a novel perspective on the user cold-start problem. Finally, we demonstrate these concepts with an empirical investigation of a state-of-the-art model trained on a widely used movie ratings dataset., Comment: appeared at FAccT '20

Time limits on benefit receipt became a central feature of federal welfare policy in the 1996 Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA). Proponents of welfare reform argued that the time limits in the Temporary Assistance for Needy Families (TANF) program, would send a firm message to recipients that welfare is intended to be temporary and that, when presented with a deadline, recipients would find jobs or other sources of support. Conversely, critics pointed out that many welfare recipients have low levels of education and skills and other personal and family challenges that make steady work difficult, and they predicted that time limits would cause harm to many vulnerable families. This report provides an examination of what has been learned to date about time limits: the implementation of state policies, the number of families affected by time limits, the effects of time limits on employment and welfare receipt, and the circumstances of families whose welfare cases have been closed because they reached a time limit. It updates a study conducted in 2002 that examined states' and families' early experiences with TANF time limits. Reported findings include: (1) States have developed varying approaches to time limits; (2) States provide exemptions from or extensions to their time limits for certain groups of families, but the policies and processes for identifying families differ from state to state; (3) Staff reported that recipients' awareness of time limits has increased over time, especially in states where many families have been terminated from assistance; (4) When TANF agencies rely on workforce agencies to deliver employment services, the time-limit message may be diluted; (5) About half of all TANF assistance cases are subject to the federal time limit; (6) Only a small portion of TANF assistance cases have at least 60 months of assistance; (7) Compared with those who have accumulated fewer months, families who have reached 60 months are headed by individuals who are older, on average; have lower levels of education; are more likely to have a disabled family member; and are more likely to be living in public housing or receiving a rent subsidy; (8) Since PRWORA was enacted, at least a quarter million cases have been closed due to reaching either a state or a federal time limit, although about one-third of the closures occurred in New York, which routinely provides post-time-limit assistance funded through an SSP; (9) Families whose benefits were terminated because of time limits were more likely than all other case closures to lack a high school education, to have never married, to be living in public housing or receiving a rent subsidy, and to be African-American; (10) There is some evidence that time limits can encourage welfare recipients to find jobs and leave welfare more quickly, even before reaching the limit; however, the magnitude of this effect is not clear; (11) It does not appear that the cancellation of welfare benefits at a time limit induces many recipients to go to work in the short term; (12) Welfare reform initiatives with time limits have generated few overall effects on family income, material hardship, or household composition in the period after families began reaching the limits, although it is not possible to isolate the effects on families whose benefits were terminated; (13) The employment rates of time-limit leavers after exiting welfare vary widely across states, ranging in these studies from less than 50 percent to more than 80 percent; (14) Many families whose benefits were terminated due to time limits continued to receive some form of public assistance after leaving TANF; and (15) Families whose benefits were terminated due to time limits reported financial struggles and, in some states, experienced higher levels of material hardships than they had while on TANF. Overall, it appears that time limits have not generated as much attention or caused as much harm to the typical family on TANF as critics of PRWORA feared. Most states have implemented stricter work participation requirements since PRWORA was enacted and tougher sanctioning policies. Perhaps as a result, families are not reaching state and federal time limits in large numbers. Little is known regarding how families who have reached time limits at later dates are faring. More research is needed that focuses on different cohorts of leavers to understand whether they are receiving other benefits (such as Medicaid and food stamps), whether they are employed, and whether they are experiencing material hardships. The report concludes that further tracking of state-made changes policies and implementation practices to determine the effect of Deficit Reduction Act time-limit policies and the outcomes of families reaching the time limit. Four appendixes are included: (1) State Policies Regarding Time Limits for the Temporary Assistance for Needy Families Program; (2) Profiles of the States Discussed; (3) Supplemental Analysis Based on Quarterly Data from the Temporary Assistance for Needy Families Program; and (4) Background Information on Welfare Leaver Studies. (Contains 47 footnotes, 3 boxes, 6 figures and 31 tables.