What happened
Staff in the Social Security Administration's Office of Appellate Operations built Insight. It uses natural language processing to flag possible errors of policy and consistency in decisions from disability hearings. The appellate office released it to its analysts in August 2017, and all hearing offices had it by September 2018. SSA's Office of the Inspector General reviewed its use from January to November 2018. It surveyed 633 staff with access to Insight and received 302 answers. About 74% agreed that it was useful and valuable when they reviewed decisions, and about 26% disagreed. Sixty-seven users wrote that it helped their reviews, and 71 thought that its flags had accuracy problems. In its report of 19 April 2019, the Inspector General also relays a study by the appellate office. Cases of Insight users were returned for quality problems 12.6% of the time, against 21.5% for non-users. The Inspector General did not test these figures. It found that SSA was not tracking whether Insight met its goals.
Sources
-
Audit report, by authors with no tie to the system.
Social Security Administration, Office of the Inspector General. The Social Security Administration's Use of Insight Software to Identify Potential Anomalies in Hearing Decisions. Audit report A-12-18-50353, 19 April 2019 (archived copy of the report at oig-files.ssa.gov).
In addition, about 74 percent agreed the program was useful and valuable when they were reviewing decisions, while about 26 percent disagreed (see Figure 2).
Details
AI system
- Workflow
-
Insight extracted information from the text of each hearing decision and flagged possible quality issues. Each flag linked users to the relevant policies so that they could check the issue.
Follow-up
- Conditions for replication
-
Users must check each flag against policy before acting on it. The flags must be accurate and consistent with the agency's disability policies.
Entry details
- Sector
-
Social security (disability appeals).
- Purpose
-
Staff in SSA's Office of Appellate Operations built Insight to flag possible errors of policy and internal consistency in decisions from disability hearings.
Not yet checked: model, release, provider, role, configuration, inputs and outputs, autonomy, human oversight, AI contribution, beneficiaries, reach, distribution and access, duration, every measurement question, every money question, ownership and control, dependencies, cross-border connections, geopolitical implications, replication attempts, availability, institutional decision, grounds, action and result, success criterion, every evidence question, observed or intended, alternative assessments, related entries, other identifiers, category mappings.
Cite this record
Peter H. Vartanian. (2026). Most staff found Insight useful for reviewing hearing decisions (AIBD-0674). Artificial Intelligence Benefits Database, edition of 31 October 2026. https://petervartanian.xyz/aibd/records/AIBD-0674.html (archived at https://doi.org/10.7910/DVN/AADSII, version 4.0)