EAGER: A New Explainable Multi-objective Learning Framework for Personalized Dietary Recommendations against Opioid Misuse and Addiction

由于阿片类药物过量死亡在过去20年中在全国范围内持续增加，打击阿片类药物危机是国家的优先事项。虽然药物辅助治疗(MAT)被认为是治疗阿片类药物滥用和成瘾的最有效的方法，但治疗过程中产生的焦虑和抑郁以及各种副作用都会引发阿片类药物的复发。除MAT外，饮食营养干预在阿片类药物滥用预防和康复中的重要性已得到证明。然而，关于如何为阿片类药物滥用和成瘾提供有效而负担得起的个性化饮食营养干预的研究还很缺乏。为了弥补这一差距，该项目的目标是设计和开发一个新的可解释的多目标学习框架，根据阿片类药物使用者的特点和情况提出个性化的饮食建议，以打击阿片类药物的滥用和成瘾，从而有助于提高国家公共健康、安全和福利。该项目的成果，包括开放源码、基准数据和开发的模型，将通过演示、出版物和媒体媒体等向公众公布和广泛分发。有人认为，抗击阿片类药物流行需要几代人的长期承诺和努力。该项目将通过学生辅导和各种K-12外联活动将研究与教育结合起来，以培训和教育后代预防和干预阿片类药物滥用和成瘾。该团队还将扩大针对妇女和代表不足的群体在计算方面的参与。通过采用新的学科视角，这一探索性高风险-高回报项目包括三个相互关联的研究组成部分，以开发一个新的可解释的多目标学习框架，以打击阿片类药物的滥用和成瘾。首先，基于Yelp等在线平台产生的饮食数据，通过解决在线饮食数据的多模态、异质性、噪声和稀疏性问题，该团队将开发用于在线阿片类药物使用者检测的新型多模式自我监督图形学习技术，以建立首个大规模、高质量、与阿片类药物使用者相关的饮食基准数据集。其次，由于阿片类药物使用者的复杂特征和情况，为他们推荐最优饮食是一个巨大的挑战，该团队将开发一种基于多跳推理的新的多目标学习算法，将多种因素(饮食偏好、营养多样性、用户特定条件)纳入到阿片类药物使用者的个性化饮食推荐中。第三，为了进一步提高推荐接受性，该团队将设计和开发一个基于推理路径的新型编解码器文本生成模型，为阿片类药物使用者提供建议食谱的文本解释。制定的框架将加快个性化饮食营养干预，以减少阿片类药物的滥用，预计将对解决这一危机产生重大影响。这项研究将推进科学理论，使信息集成和信息学领域以及公共卫生、流行病学、社会和行为科学等多学科领域受益。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

As opioid overdose deaths have continued to increase over the past two decades across the country, combating the opioid crisis is a national priority. Although medication assisted treatment (MAT) is recognized as the most effective treatment for opioid misuse and addiction, the anxiety and depression created during the treatment and various side effects can trigger opioid relapse. In addition to MAT, dietary nutrition intervention has demonstrated its importance in opioid misuse prevention and recovery. However, research on how to provide effective yet affordable personalized dietary nutrition interventions in opioid misuse and addiction is lacking. To bridge this gap, the goal of this project is to design and develop a new explainable multi-objective learning framework for personalized dietary recommendations tailored to opioid users' characteristics and circumstances to combat opioid misuse and addiction, and thus help enhance national public health, safety, and welfare. The outcomes from this project, including open-source code, benchmark data, and developed models, will be made publicly accessible and broadly distributed through demos, publications, and media presses, etc. It has been argued that combating the opioid epidemic will take long-term commitment and effort for several generations. This project will integrate research with education via student mentoring and various K-12 outreach activities to train and educate future generations in the prevention and intervention of opioid misuse and addiction. The team will also broaden participation in computing aiming at women and underrepresented groups.By engaging novel disciplinary perspectives, this exploratory high risk-high payoff project includes three interconnected research components for the development of a new explainable multi-objective learning framework to combat opioid misuse and addiction. First, based on the dietary data generated from the online platforms such as Yelp, by addressing the issues of multi-modality, heterogeneity, noise and sparseness of the online dietary data, the team will develop novel multi-modal self-supervised graph learning techniques for online opioid user detection to establish the first large-scale, high-quality, opioid-user-related dietary benchmark dataset. Second, as it is a great challenge to recommend optimal diets for opioid users due to their complex characteristics and circumstances, the team will develop a new multi-objective learning algorithm based on multi-hop reasoning to incorporate multiple factors (diet preference, nutrient diversity, user-specific condition) for personalized dietary recommendations to opioid users. Third, to further promote recommendation receptivity, the team will design and develop a novel encoder-decoder text generation model based on the reasoning paths to provide opioid users with textual explanations of suggested recipes. The developed framework will accelerate personalized dietary nutrition interventions for reducing opioid misuse, and is expected to have a significant impact on addressing this crisis. The research will advance scientific theory and benefit the information integration and informatics domain as well as multidisciplinary areas such as public health, epidemiology, and social and behavioral sciences.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.