The Evolution of Criminal History Records Systems

Imagine applying for a job that requires a background check — perhaps as a teacher, healthcare worker, private employment, or even a T-ball coach.

Further imagine attempting to exercise your civil liberties, such as obtaining a firearm. These and several other circumstances necessitate thorough vetting to ensure the safety and well-being of our society, especially regarding vulnerable populations, such as children and the elderly, However, a glaring issue arises when these background checks return incomplete or erroneous results, which can have severe implications.

For instance, if a person was arrested but not convicted, the absence of a disposition might leave an unresolved felony charge on their criminal-history record (commonly known as a RAP sheet). This can unjustly hinder their employment opportunities, and even infringe on their constitutional rights, such as the right to vote or the ability to own a firearm. Conversely, it might lead to the hiring of individuals with serious criminal backgrounds in sensitive positions, resulting in significant risks to the public.

Traditionally, criminal-history recordkeeping for about a half century was a labor-intensive, paper-based endeavor that was cumbersome, time-consuming, prone to error, and challenging to navigate, which made record queries, like criminal-background checks, incredibly difficult.

The process, known as the repository model, typically required arresting agencies, courts, and prosecutors to submit information from their portion of the criminal-justice cycle to a central agency (deemed a repository agency), which was charged with collating this information into a meaningful record of arrests and prosecution (RAP). In the early days of repository agencies, multiple carbon copies of each record were made, with one sent to the repository agency and the rest sent down the justice chain. Issues developed, like missing information, lost or unreadable forms, and long filing or data-entry delays, which became exponentially more pronounced as the criminal-data volume grew.

In the 1970s, the situation began to improve somewhat, with the advent of electronic databases. Paper records started to be input into such databases, which were easier to manage than paper and easier to extract information from when needed. Automated fingerprint-identification systems (AFIS) also were introduced in that decade, which proved to be a huge leap forward, especially regarding the capture of arrest and charging information generated at a booking event.

Fingerprint identification was pioneered in the late 1800s. Fingerprints were found to be highly reliable identifiers for two reasons: no two fingerprints are exactly alike, even in identical twins; fingerprints remain constant through a person’s lifetime. In 1910, fingerprints were used for the first time to convict a murderer, which occurred in Illinois. The verdict was appealed but eventually upheld by the state’s Supreme Court. These court cases established the reliability of fingerprint evidence.

Eventually, Live Scan technology emerged, which enabled digitization of fingerprints and facilitated electronic submission of fingerprints and arrest charges to the repository agency. In addition, electronic system interfaces revolutionized data capture and information exchange with critical state- and federal-level databases, like the FBI’s Criminal Justice Information System (CJIS). More recently, many of these systems have transitioned to technologically advanced cloud-hosted infrastructure that significantly increases utility and security.

Even though many federal and state laws still mandate use of the repository model — which makes it challenging to shift to modern, direct-query systems that could offer real-time data from the source — these advances clearly have enhanced the effectiveness of criminal-history records systems. Nevertheless, the problem of criminal records completeness and accuracy persists.

Repository agencies and their partnering public-safety and judicial agencies continue to expend large amounts of resources to complete and correct deficiencies in criminal-history information nationally. It is common for these agencies to employ several staff resources charged with the daily task of validating records accuracy and searching for missing information. This time-sensitive task is multiplied across the country throughout several repository and related agencies nationally. This expensive endeavor is further supplemented by federal grant programs designed to help repositories find missing information and address the issues that lead to missing or inaccurate criminal information.

The justice sector commonly approaches the problem of finding missing dispositions in the following three ways:

• Backfill method — This method involves obtaining copies of state court systems and other relevant databases. Through advanced techniques these records are cross-referenced to fill in missing dispositions. This large-scale approach allows for the correction of a significant volume of records in a relatively short time and at lower cost compared with the other methods.
• Bounty method — In this approach, third-party companies are hired to physically locate missing disposition records. They are paid for each verified record they retrieve. While effective, this method is slow and expensive.
• Self-service portals — Local law enforcement and courts can log into these portals to update missing information. However, due to varying levels of engagement and resource constraints, this method has enjoyed limited success.

The backfill method, which first was conceptualized two decades ago, today augments existing criminal-history records by applying machine learning (ML) and artificial intelligence (AI) to find disposition matches. This method involves teaching the system to recognize data patterns and anomalies, such as transposed names or minor discrepancies in dates, which traditional methods might miss.

A team of MCP data scientists led by Jim Pingel, MCP’s vice president and director of data and software integration solutions, pioneered a method that utilizes AI/ML to match records more accurately and efficiently. They also leveraged ML to identify clerical errors — i.e., coding disparities, typos, and misspellings — significantly accelerating and reducing the cost of the process.

In the state of Alabama, it is estimated that currently only 37 percent of arrest records in their statewide criminal-history database have a corresponding disposition record. Using this ML-assisted matching approach in a pilot project, our Alabama client realized a 41 percent improvement in disposition match rates for the sample of arrest records that was provided. Alabama is expanding this project to apply it to the full criminal-history database.

We would love the opportunity to chat with you about this approach and how it will improve your court system’s disposition match rates — so, please reach out.

Chuck Collins is MCP’s vice president of public safety. Email him at ChuckCollins@MissionCriticalPartners.com.