Introduction: Digital technology platforms offer unparalleled opportunities to reach vulnerable adolescents at scale and overcome many barriers that exist around conventional service provision. This paper describes the design and development of POD Adventures , a blended problem-solving game-based intervention for adolescents with or at risk of anxiety, depression and conduct difficulties in India. This intervention was developed as part of the PRemIum for ADolEscents (PRIDE) research programme, which aims to establish a suite of transdiagnostic psychological interventions organized around a stepped care system in Indian secondary schools. Methods and Materials: Intervention development followed a person-centered approach consisting of four iterative activities: (i) review of recent context-specific evidence on mental health needs and preferences for the target population of school-going Indian adolescents, including a multiple stakeholder analysis of school counseling priorities and pilot studies of a brief problem-solving intervention; (ii) new focus group discussions with N = 46 student participants and N = 8 service providers; (iii) co-design workshops with N = 22 student participants and N = 8 service providers; and (iv) user-testing with N = 50 student participants. Participants were aged 12–17 years and recruited from local schools in New Delhi and Goa, including a subgroup with self-identified mental health needs ( N = 6). Results: Formative data from existing primary sources, new focus groups and co-design workshops supported a blended format for delivering a brief problem-solving intervention, with counselors supporting use of a game-based app on “offline” smartphones. User-testing with prototypes identified a need for simplification of language, use of concrete examples of concepts and practice elements to enhance engagement. There were also indications that participants most valued relatability and interactivity within real-world stories with judicious support from an in-app guide. The final prototype comprised a set of interactive and gamified vignettes and a structured set of problem-solving questions to consolidate and generalize learning while encouraging real-world application. Discussion: Findings shaped the design of POD Adventures and its delivery as an open-access blended intervention for secondary school students with a felt need for psychological support, consistent with an early intervention paradigm. A randomized controlled trial is planned to evaluate processes and impacts of POD Adventures when delivered for help-seeking students in low-resource school settings.

In many real-world sequential decision making problems, the number of available actions (decisions) can vary over time. While problems like catastrophic forgetting, changing transition dynamics, changing rewards functions, etc. have been well-studied in the lifelong learning literature, the setting where the size of the action set changes remains unaddressed. In this paper, we present first steps towards developing an algorithm that autonomously adapts to an action set whose size changes over time. To tackle this open problem, we break it into two problems that can be solved iteratively: inferring the underlying, unknown, structure in the space of actions and optimizing a policy that leverages this structure. We demonstrate the efficiency of this approach on large-scale real-world lifelong learning problems.

No abstract

Many sequential decision-making systems leverage data collected using prior policies to propose a new policy. For critical applications, it is important that high-confidence guarantees on the new policy’s behavior are provided before deployment, to ensure that the policy will behave as desired. Prior works have studied high-confidence off-policy estimation of the expected return, however, high-confidence off-policy estimation of the variance of returns can be equally critical for high-risk applications. In this paper we tackle the previously open problem of estimating and bounding, with high confidence, the variance of returns from off-policy data.

Shared autonomy refers to approaches for enabling an autonomous agent to collaborate with a human with the aim of improving human performance. However, besides improving performance, it may often also be beneficial that the agent concurrently accounts for preserving the user's experience or satisfaction of collaboration. In order to address this additional goal, we examine approaches for improving the user experience by constraining the number of interventions by the autonomous agent. We propose two model-free reinforcement learning methods that can account for both hard and soft constraints on the number of interventions. We show that not only does our method outperform the existing baseline, but also eliminates the need to manually tune a black-box hyperparameter for controlling the level of assistance. We also provide an in-depth analysis of intervention scenarios in order to further illuminate system understanding.