Query your average litigation attorney about the difference between predictive coding technology and other more traditional litigation tools, and you're likely to get a wide range of responses. The fact that “predictive coding” goes by many names, including “computer-assisted review” (CAR) and “technology-assisted review” (TAR) illustrates a fundamental problem: What is predictive coding, and how is it different from other tools in the litigator's technology toolbelt?

Predictive coding is a type of technology that enables a computer to “predict” how documents should be classified based on input or “training” from human reviewers. The technology can expedite the document review process by finding key documents faster, potentially saving organizations thousands of hours of time. And in a profession where time is money, narrowing days, weeks, or even months of tedious document review into more reasonable time frames means organizations can save big bucks by keeping litigation expenditures in check.

Despite the promise of predictive coding, widespread adoption among the legal community has been slower than expected, due in part to the confusion about how it differs from other types of TAR tools that have been available for years. Unlike TAR tools that automatically extract patterns and identify relationships between documents with minimal human intervention, predictive coding requires significant reliance on humans to train and fine-tune the system through an iterative process. Some common TAR tools used in e-discovery that do not include this same level of interaction are described below:

• Keyword search: In its simplest form, a word is input into a computer system which then retrieves the documents within the collection that contain the same word. Also commonly referred to as Boolean searching, keyword search tools typically include enhanced capabilities to identify word combinations and derivatives of root words among other things.
• Concept search: Typically involves the use of linguistic and statistical algorithms to determine whether a document is responsive to a particular key word search query. The technology considers variables such as the proximity and frequency of words that appear in relationship to the keywords used. Concept search tools retrieve more documents than keyword searches because documents containing concepts related to the keyword search are retrieved in addition to documents that contain the keyword search terms used.
• Discussion threading: Utilizes algorithms to dynamically link together related documents (most commonly e-mail messages) into chronological threads that reveal entire discussions. This technology simplifies the process of identifying participants to a conversation and understanding the substance of the conversation.
• Clustering: Involves the use of linguistic algorithms that automatically organize a large collection of documents into different topical groupings based on similarity.
• Find similar: Enables the retrieval of documents related to a particular document of interest. Reviewing similar documents together can simplify the review process, provide broader context, and help increase coding accuracy.
• Near-duplicate identification: Allows reviewers to easily identify, retrieve, and code documents that are very similar but not exact duplicates. Some systems can highlight discrepancies between near-duplicate documents which makes identifying subtle differences between documents easier.

Because of the deep level of human training involved on the front end, predictive coding technology only requires humans to review a small fraction of documents, resulting in a fraction of the review costs. The process typically begins with document reviewers using a computer system to review and classify a small sample of case documents as either responsive or non-responsive. These classification decisions are simultaneously logged by the computer, and the computer uses the information gleaned from this “training set” to construct a model for distinguishing between a responsive and non-responsive document. The model is applied to the remaining documents in order to predict how they should be classified and to rank the documents by degree of responsiveness. Although predictive coding technology is often viewed as a tool for segregating responsive and non-responsive documents, the technology can also be used to classify case documents based on other criteria, such as attorney-client privilege or key issues relevant to the case.

Training the predictive coding system is an iterative process that requires attorneys and their legal teams to evaluate the accuracy of the computer's document prediction scores throughout multiple training stages. A prediction score is simply a percentage value assigned to each document that is used to rank all the documents by degree of responsiveness. If the accuracy of the computer-generated predictions is insufficient, additional training documents can be selected and reviewed to help improve the system's performance. Multiple training sets are commonly reviewed and coded until the desired performance levels are achieved. Once the desired performance levels are achieved, informed decisions can be made about which documents to produce to the requesting party.

For example, if the legal team's analysis of the computer's predictions reveals that within a population of one million documents only those with prediction scores in the 70 percent range and higher appear to be responsive, the team may elect to review and then produce only those 300,000 documents to the requesting party. The financial consequences of this approach are significant, considering a recent RAND Corporation study estimates the cost of reviewing a single gigabyte of data is approximately $18,000. That means excluding 700,000 documents (around 10 to 15 gigabytes of data, depending on file sizes) from review can save organizations significant time and money.

Predictive coding technology is still relatively new to the legal community, and the multiple names is just one cause for confusion. Adding to that is the number of competing solutions, consultants, and “experts” in the market that help perpetuate a lot of misinformation. One claim that causes confusion is the notion that predictive coding renders other TAR tools obsolete. On the contrary, predictive coding technology should be viewed as one of many different tools in the litigator's toolbelt that can and should be used independently or in combination with other tools, depending on the needs of the case. Understanding how predictive coding technology and other TAR tools can be used together as part of a defensible e-discovery process can help organizations reduce risk and cost simultaneously. Providing the industry with this basic level of understanding will help ensure that predictive coding technology and related best practices standards will evolve in a manner that is fair to all parties — ultimately expediting rather than slowing broader adoption of this promising new technology. To learn more, download a free copy of Predictive Coding for Dummies.

Query your average litigation attorney about the difference between predictive coding technology and other more traditional litigation tools, and you're likely to get a wide range of responses. The fact that “predictive coding” goes by many names, including “computer-assisted review” (CAR) and “technology-assisted review” (TAR) illustrates a fundamental problem: What is predictive coding, and how is it different from other tools in the litigator's technology toolbelt?

Predictive coding is a type of technology that enables a computer to “predict” how documents should be classified based on input or “training” from human reviewers. The technology can expedite the document review process by finding key documents faster, potentially saving organizations thousands of hours of time. And in a profession where time is money, narrowing days, weeks, or even months of tedious document review into more reasonable time frames means organizations can save big bucks by keeping litigation expenditures in check.

Despite the promise of predictive coding, widespread adoption among the legal community has been slower than expected, due in part to the confusion about how it differs from other types of TAR tools that have been available for years. Unlike TAR tools that automatically extract patterns and identify relationships between documents with minimal human intervention, predictive coding requires significant reliance on humans to train and fine-tune the system through an iterative process. Some common TAR tools used in e-discovery that do not include this same level of interaction are described below:

• Keyword search: In its simplest form, a word is input into a computer system which then retrieves the documents within the collection that contain the same word. Also commonly referred to as Boolean searching, keyword search tools typically include enhanced capabilities to identify word combinations and derivatives of root words among other things.
• Concept search: Typically involves the use of linguistic and statistical algorithms to determine whether a document is responsive to a particular key word search query. The technology considers variables such as the proximity and frequency of words that appear in relationship to the keywords used. Concept search tools retrieve more documents than keyword searches because documents containing concepts related to the keyword search are retrieved in addition to documents that contain the keyword search terms used.
• Discussion threading: Utilizes algorithms to dynamically link together related documents (most commonly e-mail messages) into chronological threads that reveal entire discussions. This technology simplifies the process of identifying participants to a conversation and understanding the substance of the conversation.
• Clustering: Involves the use of linguistic algorithms that automatically organize a large collection of documents into different topical groupings based on similarity.
• Find similar: Enables the retrieval of documents related to a particular document of interest. Reviewing similar documents together can simplify the review process, provide broader context, and help increase coding accuracy.
• Near-duplicate identification: Allows reviewers to easily identify, retrieve, and code documents that are very similar but not exact duplicates. Some systems can highlight discrepancies between near-duplicate documents which makes identifying subtle differences between documents easier.

Because of the deep level of human training involved on the front end, predictive coding technology only requires humans to review a small fraction of documents, resulting in a fraction of the review costs. The process typically begins with document reviewers using a computer system to review and classify a small sample of case documents as either responsive or non-responsive. These classification decisions are simultaneously logged by the computer, and the computer uses the information gleaned from this “training set” to construct a model for distinguishing between a responsive and non-responsive document. The model is applied to the remaining documents in order to predict how they should be classified and to rank the documents by degree of responsiveness. Although predictive coding technology is often viewed as a tool for segregating responsive and non-responsive documents, the technology can also be used to classify case documents based on other criteria, such as attorney-client privilege or key issues relevant to the case.

Training the predictive coding system is an iterative process that requires attorneys and their legal teams to evaluate the accuracy of the computer's document prediction scores throughout multiple training stages. A prediction score is simply a percentage value assigned to each document that is used to rank all the documents by degree of responsiveness. If the accuracy of the computer-generated predictions is insufficient, additional training documents can be selected and reviewed to help improve the system's performance. Multiple training sets are commonly reviewed and coded until the desired performance levels are achieved. Once the desired performance levels are achieved, informed decisions can be made about which documents to produce to the requesting party.

For example, if the legal team's analysis of the computer's predictions reveals that within a population of one million documents only those with prediction scores in the 70 percent range and higher appear to be responsive, the team may elect to review and then produce only those 300,000 documents to the requesting party. The financial consequences of this approach are significant, considering a recent RAND Corporation study estimates the cost of reviewing a single gigabyte of data is approximately $18,000. That means excluding 700,000 documents (around 10 to 15 gigabytes of data, depending on file sizes) from review can save organizations significant time and money.

Predictive coding technology is still relatively new to the legal community, and the multiple names is just one cause for confusion. Adding to that is the number of competing solutions, consultants, and “experts” in the market that help perpetuate a lot of misinformation. One claim that causes confusion is the notion that predictive coding renders other TAR tools obsolete. On the contrary, predictive coding technology should be viewed as one of many different tools in the litigator's toolbelt that can and should be used independently or in combination with other tools, depending on the needs of the case. Understanding how predictive coding technology and other TAR tools can be used together as part of a defensible e-discovery process can help organizations reduce risk and cost simultaneously. Providing the industry with this basic level of understanding will help ensure that predictive coding technology and related best practices standards will evolve in a manner that is fair to all parties — ultimately expediting rather than slowing broader adoption of this promising new technology. To learn more, download a free copy of Predictive Coding for Dummies.