top of page

Responsible AI: Ensuring Safety at AllHere


AllHere Explained #3


As edtech companies continue to integrate AI into their offerings, it is important for them to ensure it is being used responsibly. There are various concerns surrounding AI’s use in education, including student data and privacy, inappropriate content, spreading misinformation, and bias amplification.


This post, the third in the “AllHere Explained” series, helps shed light on responsible AI and describes how AllHere ensures this with Ed™, a pioneering AI-fueled, learning acceleration platform created in partnership with the Los Angeles Unified School District.


What is responsible AI?

Responsible AI is the idea of ethical and accountable AI. It is one of The Software & Information Industry Association’s (SIIA) policy priorities in 2024.  The organization supports “policies designed to foster trustworthy and responsible AI practices, including the mitigation of unintentional bias in algorithms and training data,” as well as “industry-led efforts to raise the bar on responsible AI development and use.”


AllHere’s programs were developed using research and consultations with diverse communities, stakeholders and domain experts. They follow strict processes and controls to protect student data privacy. When AllHere creates an AI-fueled chatbot for a school or district, it undergoes pre-deployment testing, risk identification and mitigation, and ongoing monitoring to demonstrate that it is safe and effective based on its intended use.


Ed™’s safe and effective system

Built on AllHere’s award-winning platform, Ed™ was created using an approach where AI capabilities such as prediction quality and the machine learning process were prioritized from the outset. Instead of treating AI as an add-on or enhancement, AllHere focused on leveraging AI technologies as the primary driver of innovation for Ed™ from the conception phase.


All along the development phases, AllHere has followed the strictest guidelines for responsible innovation and relentlessly tested Ed™. Here are 10 examples on how AllHere ensured Ed™ demonstrates responsible AI.


●       Activate and control AI in a “walled garden”: Ed™ offers a secure, real-time, multi-modal data foundation that integrates structured, unstructured, real-time/streaming, and other data into a single data lake. This data lake prioritizes data protection and long-term sustainable solutions.

●       Tested scenarios: AllHere examined and evaluated AI-driven activity and its consequences before enacting it.

●       Researched and evidence-based tools: Ed™ incorporates only the highest-quality, research, evidence-based, and district-approved tools that have been proven to be effective in meeting the needs of students, including diverse learners such as English learners and students with special needs. As the foundation for Ed™’s instruction, this carefully curated cohort of industry-leading tools goes beyond a repository approach to focusing on accelerating learning by targeting individual students’ needs, and they represent the best in engaging, interactive lessons that promote critical thinking and personalized learning and practice in the critical areas of literacy, numeracy, and science.

●       Emphasize and employ a rigorous human-in-the loop governance framework: AllHere can pinpoint when and how humans are embedded in decision-making. We conduct frequent, structured reviews and audits to assess Ed™’s systems impact, fairness, and safety, ensuring that it meets the highest standards of responsibility and transparency. Our dynamic governance model allows us to swiftly adapt to new insights, societal expectations and regulatory changes.

●       Human-centered: Ed amplifies what people do; it does not replace them. It enhances and assists educational practices at every level. AI does not replace teachers, it helps them provide personalized support to students at scale.  

●       Built with strict processes and controls: This includes protecting student data, security against adversarial attacks, and the masking of Personally Identifiable Information (PII) data throughout model development and the deployment life cycle. Additionally, Ed™ is compliant with FERPA and other data protection regulations.

●       Built with a focus on equity and mitigating bias: AllHere believes it’s important for AI to reflect the communities it serves. Ed™ is built by a diverse group of individuals with different backgrounds, experiences, races, languages and cultures to be representative of the way school districts look.

●       Informed and involved educators, families, and students: We’ve prioritized informing and involving educational constituents so they are prepared to use AI to fit specific teaching, learning, and family engagement needs.

●       Developed from consultation: Ed™ was evaluated by diverse communities, stakeholders, and domain experts to identify concerns, risks, and potential impacts of the system.


AllHere believes responsible AI is vital as a lot of research and development of Ed™ focused on addressing context and enhancing trust and safety. We’ve done deep research that focuses Ed™ in on always adapting to context (the individual learner, variabilities in instructional approaches, what assets and differences families and students have). Ed™ operates as a ‘school of one,’ making it flexible and extendable to the district to meet each student’s, family’s, and educator’s particular needs.


About this Series: This blog post is the third in a series called “AllHere Explained.” The series will explore a range of topics, including AllHere’s approach to security, privacy, AI/ML safety, education innovation, digital transformation, and more. We hope you find these blogs useful and informative; and we welcome feedback.

Comentários


bottom of page