An official website of the United States government
Many real-life scenarios require humans to make difficult trade-offs: do we always follow all the traffic rules or do we violate the speed limit in an emergency? These scenarios force us to evaluate the trade-off between collective norms and our own personal objectives. To create effective AI-human teams, we must equip AI agents with a model of how humans make trade-offs in complex, constrained environments. These agents will be able to mirror human behavior or to draw human attention to situations where decision making could be improved. To this end, we propose a novel inverse reinforcement learning (IRL) method for learning implicit hard and soft constraints from demonstrations, enabling agents to quickly adapt to new settings. In addition, learning soft constraints over states, actions, and state features allows agents to transfer this knowledge to new domains that share similar aspects.
more »
« less
A Taxonomy of Forcing Functions for Addressing Human Errors in Human-machine Interaction
A forcing function is an intervention for constraining human behavior. However, the literature describing forcing functions provides little guidance for when and how to apply forcing functions or their associated trade-offs. In this paper, we address these shortcomings by introducing a novel taxonomy of forcing functions. This taxonomy extends the previous methods in four ways. First, it identifies two levels of forcing function solidity: hard forcing functions, which explicitly enforce constraints through the system, and soft forcing functions, which convey or communicate constraints. Second, each solidity level is decomposed into specific types. Third, the taxonomy hierarchically ranks forcing function solidities and types based on trade-offs of constraint and resilience. Fourth, for hard forcing functions, our taxonomy offers formal guidance for identifying the minimally constraining intervention that will prevent a specific error from occurring. We validated the ability of our method to identify effective error interventions by applying it to systems with known errors from the literature. We then compared the solutions offered by our method to known, effective interventions. We discuss our results and offer suggestions for further developments in future research.
more »
« less
- Award ID(s):
- 2219041
- PAR ID:
- 10344838
- Date Published:
- Journal Name:
- 2021 IEEE International Conference on Systems, Man, and Cybernetics (SMC)
- Page Range / eLocation ID:
- 3134 to 3139
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Many real-life scenarios require humans to make difficult trade-offs: do we always follow all the traffic rules or do we violate the speed limit in an emergency? In general, how should we account for and balance the ethical values, safety recommendations, and societal norms, when we are trying to achieve a certain objective? To enable effective AI-human collaboration, we must equip AI agents with a model of how humans make such trade-offs in environments where there is not only a goal to be reached, but there are also ethical constraints to be considered and to possibly align with. These ethical constraints could be both deontological rules on actions that should not be performed, or also consequentialist policies that recommend avoiding reaching certain states of the world. Our purpose is to build AI agents that can mimic human behavior in these ethically constrained decision environments, with a long term research goal to use AI to help humans in making better moral judgments and actions. To this end, we propose a computational approach where competing objectives and ethical constraints are orchestrated through a method that leverages a cognitive model of human decision making, called multi-alternative decision field theory (MDFT). Using MDFT, we build an orchestrator, called MDFT-Orchestrator (MDFT-O), that is both general and flexible. We also show experimentally that MDFT-O both generates better decisions than using a heuristic that takes a weighted average of competing policies (WA-O), but also performs better in terms of mimicking human decisions as collected through Amazon Mechanical Turk (AMT). Our methodology is therefore able to faithfully model human decision in ethically constrained decision environments.more » « less
-
In this work, we study the problem of privately maximizing a submodular function in the streaming setting. Extensive work has been done on privately maximizing submodular functions in the general case when the function depends upon the private data of individuals. However, when the size of the data stream drawn from the domain of the objective function is large or arrives very fast, one must privately optimize the objective within the constraints of the streaming setting. We establish fundamental differentially private baselines for this problem and then derive better trade-offs between privacy and utility for the special case of decomposable submodular functions. A submodular function is decomposable when it can be written as a sum of submodular functions; this structure arises naturally when each summand function models the utility of an individual and the goal is to study the total utility of the whole population as in the well-known Combinatorial Public Projects Problem. Finally, we complement our theoretical analysis with experimental corroboration.more » « less
-
Abstract The effectiveness of non-pharmaceutical interventions (NPIs) during a pandemic is challenging to assess due to the multifaceted interactions between interventions and population dynamics. Significant difficulty arises from the overlapping effects of various NPIs applied to different subgroups within a population. To address this, we propose a new mathematical model that incorporates various intervention strategies, including total and partial lockdowns, school closures, and reduced interactions among specific subgroups, such as the elderly. Our model extends previous work by explicitly accounting for the quadratic nature of control costs and the interplay between overlapping controls targeting the same population segments. Using optimal control theory, we identify intervention policies that effectively mitigate disease transmission while balancing economic and societal costs. To demonstrate the utility of our approach, we apply the model to real-world data from the COVID-19 pandemic in the State of New Jersey. Our results provide insights into the trade-offs and synergies of different NPIs and the importance of accurately capturing the relationship between a policy and the population affected.more » « less
-
Innate, infection-preventing resistance often varies between host life stages. Juveniles are more resistant than adults in some species, whereas the opposite pattern is true in others. This variation cannot always be explained by prior exposure or physiological constraints and so it has been hypothesized that trade-offs with other life-history traits may be involved. However, little is known about how trade-offs between various life-history traits and resistance at different life stages affect the evolution of age-specific resistance. Here, we use a mathematical model to explore how trade-offs with natural mortality, reproduction and maturation combine to affect the evolution of resistance at different life stages. Our results show that certain combinations of trade-offs have substantial effects on whether adults or juveniles are more resistant, with trade-offs between juvenile resistance and adult reproduction inherently more costly than trade-offs involving maturation or mortality (all else being equal), resulting in consistent evolution of lower resistance at the juvenile stage even when infection causes a lifelong fecundity reduction. Our model demonstrates how the differences between patterns of age-structured resistance seen in nature may be explained by variation in the trade-offs involved and our results suggest conditions under which trade-offs tend to select for lower resistance in juveniles than adults.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- https://doi.org/10.1109/SMC52423.2021.9658721
