Machine learning models are often criticized as opaque from a lack of transparency in their decision-making process. This study examines how policy design impacts the quality of explanations in ML models. We conducted a classroom experiment with 124 participants and analyzed the effects of policy length and purpose on developer compliance with policy requirements. Our results indicate that while policy length affects engagement with some requirements, policy purpose has no effect, and explanation quality is generally poor. These findings highlight the challenge of effective policy development and the importance of addressing diverse stakeholder perspectives within explanations.

We connect a broad class of generative models through their shared reliance on sequential decision making. Motivated by this view, we develop extensions to an existing model, and then explore the idea further in the context of data imputation -- perhaps the simplest setting in which to investigate the relation between unconditional and conditional generative modelling. We formulate data imputation as an MDP and develop models capable of representing effective policies for it. We construct the models using neural networks and train them using a form of guided policy search. Our models generate predictions through an iterative process of feedback and refinement.

Scientists are scrambling to work out what effect specific measures, such as social distancing, have in slowing the spread of COVID-19.Credit: Ivan Romano/Getty Hong Kong seems to have given the world a lesson in how to effectively curb COVID-19. With a population of 7.5 million, it has reported just 4 deaths. Researchers studying Hong Kong's approach have already found that swift surveillance, quarantine and social-distancing measures, such as the use of face masks and school closures, helped to cut coronavirus transmission -- measured by the average number of people each infected person infects, or R -- to close to the critical level of 1 by early February. Working out the effectiveness of the unprecedented measures implemented worldwide to limit the spread of the coronavirus is now one of scientists' most pressing questions. Researchers hope that, ultimately, they will be able to accurately predict how adding and removing control measures affects transmission rates and infection numbers.

We connect a broad class of generative models through their shared reliance on sequential decision making. Motivated by this view, we develop extensions to an existing model, and then explore the idea further in the context of data imputation -- perhaps the simplest setting in which to investigate the relation between unconditional and conditional generative modelling. We formulate data imputation as an MDP and develop models capable of representing effective policies for it. We construct the models using neural networks and train them using a form of guided policy search. Our models generate predictions through an iterative process of feedback and refinement.

OSAKA - When times are good, there is less political pressure at the local level anywhere to be economically efficient or carefully scrutinize predictions that a new public works project or expensive industrial or tourism promotion scheme will lead to prosperity in 20 or 30 years. But with their rapidly aging and declining populations and shrinking tax bases, local governments now face a daunting task in formulating political, economic, social and environmental policies that will most likely benefit the greatest number of people decades from now. In contrast to the carefree public works spending of the bubble economy of three decades ago, often based on proposals that seemed little thought out, the demand for data-driven, evidence-based projections for various policy measures among local governments has grown, lest a wrong decision lead to local economic disaster, and voter anger. Earlier this year, Nagano Prefecture announced it would rely more on computer modeling and scenarios for local policy decisions. The decision came after the prefecture cooperated with Kyoto University's Kokoro Research Center, Hitachi Ltd. and Mitsubishi UFJ Research and Consulting to create two different models using artificial intelligence. Those were put to use in research on the best policy to realize a sustainable society and how to best take advantage of the opportunities, especially related to local tourism, that might come from the planned opening of a maglev shinkansen station in the prefecture as early as 2027.

Pieter started his invited talk by summarizing some of the key differences between supervised learning and Reinforcement Learning (RL). In essence, RL is mainly concerned with learning an effective policy to have an agent interact with the world in a way that best achieves a goal. For example, learning a policy on how to walk. Recently, RL has seen many success stories, such as learning to play Atari games from the raw pixel inputs, mastering the game of Go to a superhuman level, or effectively teaching simulated characters how to walk from scratch. However, one big gap between RL algorithms and humans, remains the time it takes to acquire new and effective policies.