Search for: All records

In settings where an AI agent nudges a human agent toward a goal, the AI can quickly learn a high-quality policy by modeling the human well. Despite behavioral evidence that humans hyperbolically discount future rewards, we model human as Markov Decision Processes (MDPs) with exponential discounting. This is because planning is difficult with non-exponential discounts. In this work, we investigate whether the performance benefits of modeling humans as hyperbolic discounters outweigh the computational costs. We focus on AI interventions that change the human's discounting (i.e. decreases the human's "nearsightedness" to help them toward distant goals). We derive a fixed exponential discount factor that can approximate hyperbolic discounting, and prove that this approximation guarantees the AI will never miss a necessary intervention. We also prove that our approximation causes fewer false positives (unnecessary interventions) than the mean hazard rate, another well-known method for approximating hyperbolic MDPs as exponential ones. Surprisingly, our experiments demonstrate that exponential approximations outperform hyperbolic ones in online learning, even when the ground-truth human MDP is hyperbolically discounted. 

Abstract Macrotextured silicone breast implants are associated with several complications, ranging from seromas and hematomas to the formation of a rare type of lymphoma, known as breast implant-associated anaplastic large cell lymphoma (BIA-ALCL). The presence of silicone wear debris has been detected within the peri-implant region and fibrotic capsule and histological analyses reveal inflammatory cells surrounding debris particles. However, it is unclear how these debris particles are generated and released from macrotextured implant surfaces, and whether wear debris generation is related to implant stiffness. In this study, we created an accelerated implant aging model to investigate the formation of silicone wear debris produced from self-mated (“shell-shell”) tribological interactions. We created implant-like silicone elastomers from polydimethylsiloxane (PDMS) using Sylgard 184 base:curing agent (10:1, 12:1, and 16:1) and quantified their mechanical properties (E* = 1141 ± 472, 336 ± 20, and 167 ± 53 kPa, respectively). We created macrotextured PDMS samples using the lost-salt technique and compared their self-mated friction coefficient (< µ > = 4.8 ± 3.2, 4.9 ± 1.8, and 6.0 ± 2.3, respectively) and frictional shear stress (τ = 3.1 ± 1.3, 3.2 ± 1.7, and 2.4 ± 1.4 MPa, respectively) to those of the recalled Allergan Biocell macrotextured implant shell (E* = 299 ± 8 kPa, < µ > = 2.2, andτ = 0.8 ± 0.1). Friction coefficient and frictional shear stress were largely insensitive to variations in elastic modulus for macrotextured PDMS samples and recalled implant shells. The stiffest 10:1 PDMS macrotextured sample and the recalled implant shell both generated similar area fractions of silicone wear debris. However, the recalled implant shell released far more particles (> 10×), mainly within the range of 5 to 20 µm²in area. Bone marrow-derived macrophages (BMDMs) were treated with several concentrations of tribologically generated silicone wear debris. We observed widespread phagocytosis of wear debris particles and increasing secretion of inflammatory cytokines with increasing concentration of wear debris particles. Our investigation highlights the importance of avoiding macrotextured surfaces and mitigating wear debris generation from silicone implants to reduce chronic inflammation.