reSee.it Video Transcript AI Summary
Linda McLaughlin and her colleagues present a data-focused argument alleging election fraud in Georgia, supported by multiple data analyses and demonstrations.
- Linda McLaughlin introduces the data integrity group and states that data is numerical and non-partisan; she aims to remedy a lack of presented data in the discussion.
- Dave Labou, a lead data scientist, explains that their analysis across precincts, counties, and the state identified over 40 data points of negative voting or vote switching across candidates totaling over 200,000 votes. Separately, machine learning algorithms used for anomaly detection in fraud detection flagged over 500 precincts with over 1,000,000 corresponding votes showing suspicious activity. He emphasizes that the process is scientific and not tied to political affiliations.
- Labou uses a banking analogy to illustrate data integrity concerns: in hypothetical online banking, deposits or withdrawals being redirected or split would indicate fraudulent activity. He applies this concept to voting data, arguing that the voting system data aligns with the Secretary of State data used to certify results, yet exhibits patterns akin to transfers and reallocation not authorized by voters.
- He states that the data are publicly available but require advanced programming to extract, parse, and join datasets. Their independent team has made all analysis, programs, and data public to allow replication and has produced videos to translate the analysis for broader understanding.
- A key claim is that receiving over 90% in a precinct is a marker for fraud; in Fulton County, more than 150 precincts voted 90% or more for Biden, and in the statewide race (decided by less than 13,000 votes), these 150 Fulton precincts accounted for 152,000 Biden votes, described as a clear indicator of suspicious or fraudulent activity.
- Labou and team present a series of visuals and explanations indicating explicit vote count switching, e.g., in Dodge County, where Trump’s votes appear to be subtracted while Biden’s counts increase in tandem with county updates, leading to a shift in totals that would not appear in state totals due to timing of updates.
- They reference adjudication as the review of ballots flagged during scanning, noting that only ballots with a contest causing questions about how the computer reads them are adjudicated.
- In DeKalb County, they assert it is statistically impossible for nine out of ten voters to vote for Biden in 94 precincts.
- They describe a data flow in Fulton County: poll pad check-in, ballot image saved on the machine, SD cards transported to drop-off locations, escorted to a warehouse, run through Democracy Suite, exported to a Dominion server, and inserted into a SQL Server database before transmission to the Secretary of State and data aggregators.
- A critical point is the vulnerability within the county update data-entry process: the square box detailing data-entry options in the election software allows updating vote batches, projecting batches, and generating new or temporary batches that can be injected directly into the tally; these options can be validated and published, enabling potential manipulation before server upload.
- They pose questions about validation: whether two observers from both parties were present during SD card transmissions and drop-off transmissions, and whether there is a public log of exchanges at drop-off points. They challenge why elected officials have not pursued these questions about voting integrity.
- Labou notes the process is machine-to-machine and, by design, should not decrement sums; any decrement requires a robust explanation, and their data suggest negative drops are inconsistent with normal sequential processes.
Speaker 2 clarifies the data sources (CITL election night data and Edison/New York Times data) and asserts that the process from poll pads to secretary of state is machine-driven, with no human entry of totals, thereby removing human entry error as an explanation for observed negative changes. Speaker 4 adds emphasis on the validation and potential vulnerabilities in the software options used for election administration, underscoring the need for transparency and inquiry into the electoral process.