Helen Jiang

Helen Jiang, Erwen Senge

Explainable AI (XAI) research has been booming, but the question "$\textbf{To whom}$ are we making AI explainable?" is yet to gain sufficient attention. Not much of XAI is comprehensible to non-AI experts, who nonetheless, are the primary audience and major stakeholders of deployed AI systems in practice. The gap is glaring: what is considered "explained" to AI-experts versus non-experts are very different in practical scenarios. Hence, this gap produced two distinct cultures of expectations, goals, and forms of XAI in real-life AI deployments. We advocate that it is critical to develop XAI methods for non-technical audiences. We then present a real-life case study, where AI experts provided non-technical explanations of AI decisions to non-technical stakeholders, and completed a successful deployment in a highly regulated industry. We then synthesize lessons learned from the case, and share a list of suggestions for AI experts to consider when explaining AI decisions to non-technical stakeholders.

Helen Jiang, Erwen Senge

While the applications and demands of Machine learning (ML) systems in mental health are growing, there is little discussion nor consensus regarding a uniquely challenging aspect: building security methods and requirements into these ML systems, and keep the ML system usable for end-users. This question of usable security is very important, because the lack of consideration in either security or usability would hinder large-scale user adoption and active usage of ML systems in mental health applications. In this short paper, we introduce a framework of four pillars, and a set of desired properties which can be used to systematically guide and evaluate security-related designs, implementations, and deployments of ML systems for mental health. We aim to weave together threads from different domains, incorporate existing views, and propose new principles and requirements, in an effort to lay out a clear framework where criteria and expectations are established, and are used to make security mechanisms usable for end-users of those ML systems in mental health. Together with this framework, we present several concrete scenarios where different usable security cases and profiles in ML-systems in mental health applications are examined and evaluated.

Helen Jiang

A mental health crisis is looming large, and needs to be addressed. But across age groups, even just in the United States, more than 50% of people with any mental illness (AMI) did not seek or receive any service or treatment. The proliferation of telehealth and telepsychiatry tools and systems can help address this crisis, but outside of traditional regulatory aspects on privacy, e.g. Health Insurance Portability and Accountability Act (HIPPA), there does not seem to be enough attention on the security needs, concerns, or user experience of people with AMI using those telehealth systems. In this text, I try to explore some priority security properties for telehealth systems used by people with AMI for mental heath services (MHS). I will also suggest some key steps in a proposed process for designing and building security mechanisms into such systems, so that security is accessible and usable to patients with AMI, and these systems can achieve their goals of ameliorate this mental health crisis.

Devendra Singh Chaplot, Helen Jiang, Saurabh Gupta et al.

In this paper, we study the task of embodied interactive learning for object detection. Given a set of environments (and some labeling budget), our goal is to learn an object detector by having an agent select what data to obtain labels for. How should an exploration policy decide which trajectory should be labeled? One possibility is to use a trained object detector's failure cases as an external reward. However, this will require labeling millions of frames required for training RL policies, which is infeasible. Instead, we explore a self-supervised approach for training our exploration policy by introducing a notion of semantic curiosity. Our semantic curiosity policy is based on a simple observation -- the detection outputs should be consistent. Therefore, our semantic curiosity rewards trajectories with inconsistent labeling behavior and encourages the exploration policy to explore such areas. The exploration policy trained via semantic curiosity generalizes to novel scenes and helps train an object detector that outperforms baselines trained with other possible alternatives such as random exploration, prediction-error curiosity, and coverage-maximizing exploration.

Nikos Papasarantopoulos, Helen Jiang, Shay B. Cohen

We describe a technique for structured prediction, based on canonical correlation analysis. Our learning algorithm finds two projections for the input and the output spaces that aim at projecting a given input and its correct output into points close to each other. We demonstrate our technique on a language-vision problem, namely the problem of giving a textual description to an "abstract scene".