Earlier this month, the White House Office of Science and Technology Policy (OSTP) Scientific Integrity Task Force released a report titled “Protecting the Integrity of Government Science.” While broad-based and overarching, the report calls attention to the particular challenges associated with artificial intelligence and machine learning.
The OSTP report is a response to President Biden’s directive to ensure lawmakers are making decisions based on scientific evidence and data with the aim of restoring the public’s trust in the government.
The document identifies scientific integrity principles and best practices meant to verify that “science is conducted, managed, communicated, and used in ways that preserve its accuracy and objectivity and protect it from suppression, manipulation, and inappropriate influence—including political interference.”
The foreword to the report — signed by OSTP Director Eric Lander, OSTP Deputy Director for Science and Society Alondra Nelson and OSTP Deputy Director for Climate and Environment Jane Lubchenco — lists the five new principles that will guide the agency’s scientific integrity policy:
- Dissent. Science benefits from dissent within the scientific community to sharpen ideas and thinking. Scientists’ ability to freely voice the legitimate disagreement that improves science should not be constrained.
- Whole of Government. Because evidence-based policymaking happens across government, scientific integrity policies should apply not only to “science agencies,” but to all Federal agencies and departments engaged in the production, analysis, communication, and use of evidence, science, and technology. These policies must also apply to all career employees, contractors, and political appointees.
- Science at the policy table. For science to inform policy and management decisions, it needs to be understood and actively considered during decision-making. This requires having scientists participate actively in policy-making.
- Transparency in sharing science. Transparency underpins the robust generation of knowledge and promotes accountability to the American public. Federal scientists should be able to speak freely, if they wish, about their unclassified research, including to members of the press.
- Accountability. Violations of scientific integrity should be considered on par with violations of government ethics, with comparable consequences.
These build on the original six principles set out by the Obama administration[i] in 2009.
The report cites additional key themes that scientific integrity policies should address, such as improving diversity, equity, inclusion, and accessibility (DEIA) and ameliorating potential bias in artificial intelligence and machine learning.
The report acknowledges: “AI and ML algorithms can magnify biases inherent in underlying data source and may contain their own inherent biases that lead to inaccurate findings, conclusions, and policy decisions. Lack of transparency into ML algorithms can undermine trust in the outcomes generated and ultimately in science and government. The concentration of data and AI capabilities in the hands of the Federal Government and private sector organizations may create inequities in who can conduct leading-edge research and who can access and make use of the results of such work.”
The task force advises that scientific integrity policies should help ensure that:
- AI and ML do not magnify biases inherent in the data they analyze or are trained on.
- Transparency is provided into ML algorithms.
- Quality of data used for AI and ML, including in their generation, sharing, and use.
- Privacy considerations are incorporated into AI and ML processes and privacy risks are mitigated.
- Transparency and access are provided to data and AI capabilities.
OSTP is planning to use the report’s data to create a near-term plan for assessing and enhancing scientific-integrity policies that agency leaders can deploy within their organizations.
The Scientific Integrity Task Force formed in May 2021 and includes 57 members from 29 agencies. The task force “addresses short-term, high-priority actions to strengthen scientific integrity and also lays the groundwork for longer-term coordination of Federal agency scientific integrity efforts.”
[i] The original six principles are:
- The selection and retention of candidates for science and technology positions in the Executive Branch should be based on the candidate’s knowledge, credentials, experience, and integrity.
- Each agency should have appropriate rules and procedures to ensure the integrity of the scientific process within the agency.
- When scientific or technological information is considered in policy decisions, the information should be subject to well-established scientific processes, including peer review where appropriate, and each agency should appropriately and accurately reflect that information in complying with and applying relevant statutory standards.
- Except for information that is properly restricted from disclosure under procedures established in accordance with statute, regulation, Executive Order, or Presidential Memorandum, each agency should make available to the public the scientific or technological findings or conclusions considered or relied on in policy decisions.
- Each agency should have in place procedures to identify and address instances in which the scientific process or the integrity of scientific and technological information may be compromised.
- Each agency should adopt such additional procedures, including any appropriate whistleblower protections, as are necessary to ensure the integrity of scientific and technological information and processes on which the agency relies in its decision-making or otherwise uses or prepares.