SBIR Phase I: An inclusive machine learning-based digital platform to credential soft skills

SBIR 第一阶段：一个基于机器学习的包容性数字平台，用于认证软技能

• 批准号: 2317077
• 负责人: Geeta Verma
• 金额: $ 27.5万
• 依托单位: LIVEDX INC.
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2024
• 资助国家: 美国
• 起止时间: 2024-01-01 至 2024-06-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2317077&HistoricalAwards=false
• 关键词: SBIR; Phase; inclusive; machine; learning

The broader/commercial impact of this Small Business Innovation Research (SBIR) Phase I project is to enable people who aspire to higher education and/or career opportunities to create a demonstrable portfolio of soft skills based on their lived experiences. Soft skills (e.g., problem-solving, teamwork, leadership, etc.) are as important as hard skills for individual success. However, current soft-skill assessment tools are subjective, inefficient, and inconsistent. This is especially painful for marginalized populations such as minorities and women, who often possess valuable soft skills such as stress management and conflict resolution, but do not have the tools to demonstrate it. The proposed solution will change how people’s lived experiences and the soft skills associated to those experiences are valorized. This technology may open the door to better educational and professional opportunities in the U.S., to increased economic competitiveness (since higher education plays an increasingly critical role in the economic competitiveness of a nation), to advanced health and welfare of the American public (since adults with higher education often live healthier and longer lives, and enjoy better financial situations), and to a more developed and diverse STEM workforce (by focusing on valorizing the social and cultural capital of minoritized students).This project proposes a digital platform that provides soft-skill credentialing guided by lived experiences. The main innovation behind the proposed solution is a proprietary system that combines Machine Learning (ML) and Natural Language Processing to analyze the candidate’s experiences and apply different evidence-based social-emotional assessment frameworks to accredit the soft skills embedded in each experience. This solution may be the first time a proprietary ML technology will be integrated with a large language model to provide soft-skill credentialing upon lived experiences. The main technical challenge is avoiding bias in the assignation of soft-skill credentials. Other technical challenges are: 1) the potential scarcity of training data; 2) the correct definition of credential categories; and 3) the ability to explain the ML models. This project is intended to address these challenges by 1) developing a proof-of-concept prototype of the accreditation model; 2) conducting a preliminary analysis of its fairness when assessing marginalized groups; 3) reformulating the accreditation algorithm in case any bias is detected; and 4) evaluating, with real datasets, the performance of the credential classifier, the bias mitigation strategies, and the explanations generated for each assessment.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.

小型企业创新研究(SBIR)第一阶段项目的更广泛/商业影响是使渴望接受高等教育和/或职业机会的人能够根据他们的生活经验创建一套可展示的软技能组合。软技能(例如，解决问题、团队合作、领导力等)对于个人的成功来说，这些技能与艰苦的技能同样重要。然而，目前的软技能评估工具是主观的、低效的、不一致的。这对少数群体和妇女等边缘化群体尤其痛苦，他们往往拥有压力管理和解决冲突等宝贵的软技能，但没有工具来证明这一点。拟议的解决方案将改变人们的生活体验以及与这些体验相关的软技能的评估方式。这项技术可能会为美国提供更好的教育和职业机会，提高经济竞争力(因为高等教育在一个国家的经济竞争力中发挥着越来越重要的作用)，为提高美国公众的健康和福利(因为受过高等教育的成年人通常更健康、更长寿，享受更好的财务状况)，以及为更发达和多样化的STEM劳动力(通过对小学生的社会和文化资本进行估价)打开大门。该项目提出了一个数字平台，提供基于生活经验的软技能认证。建议解决方案背后的主要创新是一个专有系统，该系统结合了机器学习(ML)和自然语言处理来分析应聘者的经历，并应用不同的基于证据的社会情绪评估框架来评估每种经历中嵌入的软技能。这个解决方案可能是第一次将专有的ML技术与大型语言模型相集成，以提供基于生活经验的软技能认证。主要的技术挑战是避免软技能证书分配中的偏见。其他技术挑战是：1)训练数据的潜在稀缺性；2)证书类别的正确定义；以及3)解释ML模型的能力。该项目旨在通过以下方式应对这些挑战：1)开发认证模型的概念验证原型；2)在评估边缘群体时对其公平性进行初步分析；3)在检测到任何偏见的情况下重新制定认证算法；以及4)使用真实数据集评估证书分类器的性能、缓解偏见的策略以及每次评估产生的解释。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。