Logol, our Exhibition Sponsor for the 7th World Legal Tech Summit, shares the range of error accepted in legal AI vs the errors humans make. Continue reading:

The increasing adoption of Artificial Intelligence (AI) in the legal sector has sparked numerous debates, ranging from ethical considerations to efficacy. One vital aspect often scrutinized is the “range of error” tolerated when AI is deployed alongside human lawyers. This article aims to explore the scope of acceptable errors for both AI and humans in the legal profession and contrast these ranges to identify advantages and limitations.

Defining “Range of Error” in the Legal Sector

In legal parlance, the “range of error” refers to the accepted level of inaccuracies or mistakes made during professional activities such as research, contract analysis, or even court representation. This range is often undefined for human professionals, as errors can result in disciplinary actions, lawsuits, or professional humiliation but are sometimes considered inevitable due to human fallibility.

For AI, the range of error is usually expressed as a percentage rate calculated during testing phases on sample data sets. An acceptable error rate would be defined based on the specific use case and could vary widely.

Legal AI Error Range

When AI systems are implemented in legal firms for tasks like document review, case prediction, or legal research, their error rates are rigorously evaluated. Depending on the application, a typical “good performance” might entail an error rate of less than 5%. For instance, in e-discovery tools designed for document review, an error rate of 2-4% might be considered acceptable.

Human Error in the Legal Sector

Human error in the legal profession is less quantifiable but can be equally significant. For example, oversight in a critical contract clause could lead to million-dollar losses. Past studies have suggested that the error rates among human paralegals and junior lawyers in tasks like document review could range from 10-20%.

Comparative Advantages and Disadvantages

Speed and Volume

AI can process a higher volume of work at a speed no human can match. With lower error rates in data-heavy tasks like e-discovery or contract analysis, AI has a clear edge over human capability.

Complex Judgments

Humans excel in tasks requiring emotional intelligence, complex reasoning, and ethical judgments. In these areas, AI’s “range of error” is considerably higher, as these are aspects that machine learning models currently struggle to grasp.

Ethical and Legal Accountability

Errors committed by AI systems can raise questions about accountability that do not have easy answers. In contrast, a human lawyer’s error can be attributed directly, and disciplinary actions can be taken.

Cost

AI systems involve high initial costs but can be more cost-efficient in the long run, especially when error-related costs are factored in. Humans, though expensive, bring the advantage of adaptability and complex decision-making.

Conclusion

While the range of acceptable error varies for AI and human activities in the legal sector, each has its unique advantages and disadvantages. AI systems generally have a lower error rate for high-volume, data-intensive tasks. However, they cannot yet match human performance in activities requiring complex judgments and ethical reasoning. As AI continues to evolve, a collaborative model where both humans and AI work in tandem might offer the most balanced approach, leveraging the strengths of each while minimizing their error rates.

Written by the Logol Team | Aug 28, 2023

To learn more about Logol’s state-of-the-art law practice management AI, ELLE, visit: https://elle.lawyer/