When consequential decisions are informed by algorithmic input, individuals may feel compelled to alter their behavior in order to gain a system's approval. Previous models of agent responsiveness, termed "strategic manipulation," have analyzed the interaction between a learner and agents in a world where all agents are equally able to manipulate their features in an attempt to "trick" a published classifier. In cases of real world classification, however, an agent's ability to adapt to an algorithm, is not simply a function of her personal interest in receiving a positive classification, but is bound up in a complex web of social factors that affect her ability to pursue certain action responses. In this paper, we adapt models of strategic manipulation to better capture dynamics that may arise in a setting of social inequality wherein candidate groups face different costs to manipulation. We find that whenever one group's costs are higher than the other's, the learner's equilibrium strategy exhibits an inequality-reinforcing phenomenon wherein the learner erroneously admits some members of the advantaged group, while erroneously excluding some members of the disadvantaged group. We also consider the effects of potential interventions in which a learner can subsidize members of the disadvantaged group, lowering their costs in order to improve her own classification performance. Here we encounter a paradoxical result: there exist cases in which providing a subsidy improves only the learner's utility while actually making both candidate groups worse-off--even the group receiving the subsidy. Our results reveal the potentially adverse social ramifications of deploying tools that attempt to evaluate an individual's "quality" when agents' capacities to adaptively respond differ.

A challenge in creating a dataset for machine reading comprehension (MRC) is to collect questions that require a sophisticated understanding of language to answer beyond using superficial cues. In this work, we investigate what makes questions easier across recent 12 MRC datasets with three question styles (answer extraction, description, and multiple choice). We propose to employ simple heuristics to split each dataset into easy and hard subsets and examine the performance of two baseline models for each of the subsets. We then manually annotate questions sampled from each subset with both validity and requisite reasoning skills to investigate which skills explain the difference between easy and hard questions. From this study, we observed that (i) the baseline performances for the hard subsets remarkably degrade compared to those of entire datasets, (ii) hard questions require knowledge inference and multiple-sentence reasoning in comparison with easy questions, and (iii) multiple-choice questions tend to require a broader range of reasoning skills than answer extraction and description questions. These results suggest that one might overestimate recent advances in MRC.

Two new $1 million awards from the National Science Foundation aim to help researchers take advantage of the latest advances in data science, networking and computation while protecting the integrity of their scientific work. The Renaissance Computing Institute (RENCI) of the University of North Carolina at Chapel Hill will serve as lead institution on both projects. Many scientists today use sophisticated data-intensive approaches to combine and analyze large data sets from scientific instruments and data stores all over the country. While these techniques hold great value for discovery and innovation, integrating the necessary data and tools into a scientist's workflow is often a complex undertaking. In addition, errors can be introduced when data is moved or analyzed; if those errors go undetected, it can compromise the science.

IBM Research and Rensselaer Polytechnic Institute (RPI) are collaborating on a new approach to help students learn Mandarin. The strategy pairs an AI-powered assistant with an immersive classroom environment that has not been used previously for language instruction. The classroom, called the Cognitive Immersive Room (CIR), makes students feel as though they are in restaurant in China, a garden, or a Tai Chi class, where they can practice speaking Mandarin with an AI chat agent. The CIR was developed by the Cognitive and Immersive Systems Lab (CISL), a research collaboration between IBM Research and RPI. When learning a new language, especially one as difficult as Mandarin, it's important that students have many opportunities to speak and practice their conversational skills.

Tyson Navarro, 10, of Fremont, Calif., learns to build code using an iPad at a youth workshop at the Apple store in 2013. Code.org said a record number of female and under-represented minority students took AP computer science classes in 2018. SAN FRANCISCO -- Female, black and Latino students took Advanced Placement computer science courses in record numbers, and rural student participation surged this year, as the College Board attracted more students to an introductory course designed to expand who has access to sought-after tech skills. This year, 135,992 students took advanced placement (AP) computer science exams, a 31 percent increase from last year, according to data from the College Board, the organization that administers standardized tests that help determine college entrances as well as AP courses. Females and under-represented minorities were among the fastest growing groups.

With the tremendous generation of data, there are more and more situations where observations are described by more than one view. This is for example the case with multilingual documents that convey the same information in different languages or images that are naturally described according to different set of features (for example SIFT, HOG, CNN etc). In this paper, we study the related machine learning problem that consists in finding an efficient classification model from different information sources that describe the observations. This topic, called multiview learning Atrey et al. [2010], Sun [2013], has been expanding over the past decade, spurred by the seminal work of Blum and Mitchell on co-training Blum and Mitchell [1998] (with only two views). The aim is to learn a classifier which performs better than classifiers trained over each view separately (called view-specific classifier).

Cloze tests are widely adopted in language exams to evaluate students' language proficiency. In this paper, we propose the first large-scale human-created cloze test dataset CLOTH, containing questions used in middle-school and high-school language exams. With missing blanks carefully created by teachers and candidate choices purposely designed to be nuanced, CLOTH requires a deeper language understanding and a wider attention span than previously automatically-generated cloze datasets. We test the performance of dedicatedly designed baseline models including a language model trained on the One Billion Word Corpus and show humans outperform them by a significant margin. We investigate the source of the performance gap, trace model deficiencies to some distinct properties of CLOTH, and identify the limited ability of comprehending the long-term context to be the key bottleneck.

We present QuAC, a dataset for Question Answering in Context that contains 14K information-seeking QA dialogs (100K questions in total). The dialogs involve two crowd workers: (1) a student who poses a sequence of freeform questions to learn as much as possible about a hidden Wikipedia text, and (2) a teacher who answers the questions by providing short excerpts from the text. QuAC introduces challenges not found in existing machine comprehension datasets: its questions are often more open-ended, unanswerable, or only meaningful within the dialog context, as we show in a detailed qualitative evaluation. We also report results for a number of reference models, including a recently state-of-the-art reading comprehension architecture extended to model dialog context. Our best model underperforms humans by 20 F1, suggesting that there is significant room for future work on this data. Dataset, baseline, and leaderboard available at http://quac.ai.

As machine learning algorithms are increasing used in security-critical applications, there is a growing need to design them with active adversaries in mind. A class of adversarial attacks on machine learning that have received much attention is data poisoning attacks [21, 20, 19, 4, 7, 11]. Here, an adversary is aware of the learner's training data and algorithm, and has the power to alter a small fraction of the training data in order to make the trained classifier satisfy certain objectives. For example, a sabotage adversary may try to degrade the overall accuracy of the trained classifier as part of an industrial sabotage campaign, or a profit-oriented adversary may try to poison the training data so that the resulting model favors it - say, by recommending the its products over others. While there has been a long line of prior work on data poisoning [21, 20, 19, 4, 7, 11, 3, 8], most of it has focussed in the offline setting, where a classifier or some other model is trained on a fixed input.

As news trickeld out about the mass shooting at a video game tournament in Jacksonville, Florida on Sunday, some quickly recalled that right in that same spot there was a pro-gun control event a month ago. Survivors of the February school shooting in Parkland, Florida, which killed 17 people, held an event at the Jacksonville Landing in late July. The Jacksonville Landing is a waterfront commercial district in downtown Jacksonville with lots of bars and restaurants. Our hearts are with you Jacksonville Landing. It's crazy to think that March for Our Lives Road to Change had an event there just about a month ago.