The Department of Defense, through DARPA, is seeking proposals for the Small Business Innovation Research (SBIR) program focused on "Mitigating Explicit and Implicit Bias Through Hybrid AI." The objective is to develop innovative neuro-symbolic methods that address both explicit biases, defined through natural language, and implicit biases that arise from data and model parameters in artificial intelligence systems. This initiative is critical as it aims to enhance the safety and reliability of generative AI technologies, which are increasingly recognized for their potential to perpetuate biases, thereby impacting both government and civilian applications. Proposals are due by February 26, 2025, with the opportunity opening on January 29, 2025, and further details can be found at the official solicitation link: https://www.dodsbirsttr.mil/topics-app/.

## Description

What constitutes bias can be explicitly defined using natural language (e.g., perpetuating explicitly defined harmful stereotypes), but there are also implicit biases that manifest in the distribution of data, model weights, or statistical patterns. The combination of these two modalities holds promise for countering bias, particularly in the context of generative AI. To that end, we seek submissions that combine symbolic and neural methods to counter explicit and implicit bias, respectively, in AI models while preserving accuracy.&nbsp;The use of neuro-symbolic methods in generative AI has garnered interest in recent years, particularly for improving safety. For example, the control of LLMs has been explored through the perspective of probabilistic flow circuits (a symbolic structure) that are smoothened into a differentiable form to be integrated into the loss function of LLMs for improved control after re-training [1]. On the other hand, the enforcement of constraints in generated text has also been explored through neuro-symbolic methods via grammar-constrained decoding of tokens, where the grammar takes the form of a deterministic finite automaton (a symbolic structure) [2]. The foregoing are promising use cases for neuro-symbolic AI in addressing safety and other constraints. We invite prospective performers to propose neuro-symbolic AI methods in the broader context of mitigating bias.&nbsp;

## Objective

To develop new neuro-symbolic methods that can mitigate both explicit (e.g., symbolic descriptions of bias) and implicit (e.g., the learned model parameters) biases in artificial intelligence (AI).

## Phase I

The goal of Phase I proposals is to present a new technology to address AI bias as described previously. The technology need not be mature by the end of the phase, but a convincing proof-of-concept for its utility must be demonstrated. This proof-of-concept may come in the form of a live demo, publications in peer-reviewed venues, and open-source software, among others. Phase I deliverables and milestones for this STTR should include: <ul><li>&emsp;Month 3 : report detailing technical progress made to date and tasks accomplished. </li><li>&emsp;Month 6: finalize the technical report, including remaining challenges directions to be addressed, a tentative plan for future work, and lessons learned. </li></ul>

## Phase II

Develop, install, integrate and demonstrate a prototype system determined to be the most feasible solution during the Phase I feasibility study. This demonstration should focus specifically on: <ol><li>&emsp;Validating the product-market fit between the proposed solution and the proposed topic and define a clear and immediately actionable plan for running a trial with the proposed solution and the proposed customer.</li><li>&emsp;Evaluating the proposed solution against the objectives and measurable key results as defined in the Phase I feasibility study.3.&emsp;Describing in detail how the solution can be scaled to be adopted widely (e.g., how can it be modified for scale).4.&emsp;A clear transition path for the proposed solution that takes into account input from all affected stakeholders including, but not limited to: end users, engineering, sustainment, contracting, finance, legal, and cyber security.5.&emsp;Specific details about how the solution can integrate with other current and potential future solutions.6.&emsp;How the solution can be sustainable (i.e. supportability).7.&emsp;Clearly identifying other specific DoD or governmental customers who want to use the solution.</li></ol>

## Phase III: Dual Use Applications

The contractor will pursue commercialization of the various technologies developed in Phase II for transitioning expanded mission capability to a broad range of potential government and civilian users and alternate mission applications. Interested government end users may include the Air Force, the DoD Chief Digital and AI Office (CDAO), DARPA, White House Office of Science and Tech Policy (OSTP), Dept of Education, Dept of Commerce, and NIST, all of whom have been looking at the problem of detecting and mitigating bias in AI as part of an inter-agency working group. For example, mitigating bias in one of the DoD&rsquo;s responsible AI principles and it is widely recognized that bias remains a hurdle for responsible AI adoption. Bias also remains a hurdle for operational AI adoption, to ensure robustness of AI to rare and unlikely events. Of course, these problems also affect and are pervasive in industry, thereby motivating the dual use of the proposed technologies. Example industrial applications include the de-biasing of generative models, which have been shown to both reflect inherent racial biases, but also to create new biases as a result of current de-biasing techniques.Direct access with end users and government customers will be provided with opportunities to receive Phase III awards for providing the government additional research and development, or direct procurement of products and services developed in coordination with the program.

PDF.html

Please provide the text you would like summarized.

Topic

Agency

Program

Topic Link

Additional Information