Neural networks: the power of data vs the risks of uncertainty
Neural networks trained on large amounts of data are capable of:
- Analyzing judicial practice;
- Predict the outcome of court proceedings;
- Generate document templates based on identified patterns;
- To find contradictions in the texts of normative acts.
However, neural networks have serious drawbacks.:
1. The "Black Box".
It is impossible to explain exactly how the model came to this or that conclusion. For example, why she recommended a specific article of the Civil Code of the Russian Federation. In law, where every detail matters, this is critical.
2. Learning errors.
If a neural network is trained on outdated laws or incorrect use cases, its recommendations may be erroneous. For example, councils on tax disputes may not take into account the latest changes in the Tax Code of the Russian Federation.
3. Dependence on data.
In Russia, many court decisions are not digitized, and the available data sets are often unbalanced (for example, skewed towards arbitration cases).
Expert systems: transparency and control
Expert systems operate based on pre-defined logical rules that are created by lawyers on low-code platforms such as Botman.one, without the need for programming.
Examples of tasks that expert systems solve:
- Automation of the preparation of claims, taking into account Articles 131-132 of the Civil Procedure Code of the Russian Federation;
- Verification of contracts for compliance with the requirements of the Federal Law "On Consumer Rights Protection";
- Calculation of the limitation period under Articles 196-204 of the Civil Code of the Russian Federation.
Advantages of expert systems for the legal field:
1. Transparency.
Each decision of the system can be traced back to a specific rule or rule of law. This corresponds to the principle of "explicable AI", which is critically important for courts and regulators.
2. Security.
The algorithm does not "invent" solutions — it strictly follows the inherent logic. For example, when checking an agreement, the ES will not offer a condition that contradicts Article 422 of the Civil Code of the Russian Federation on the invalidity of transactions.
3. Ease of adaptation.
When legislation changes (for example, amendments to the Federal Law "On Bankruptcy"), the rules of the system can be quickly updated without retraining the model from scratch.
Why do expert systems benefit?
1. No need for large amounts of data for training.
Training of expert systems does not require processing large amounts of confidential information, unlike neural networks that feed on Big Data.
2. Quick adjustment of algorithms.
If the ES-based algorithm makes a mistake, the lawyer will easily find the reason for it in the rules. With a neural network, you will have to analyze millions of model parameters.
3. Regulatory requirements.
Government agencies are wary of "opaque" algorithms.
Expert systems are easily audited and comply with regulatory requirements.
Where are neural networks still relevant?
Neural networks can complement expert systems in some tasks:
- Text analysis (for example, assessing the risk of challenging a contract);
- Classification of documents.
Expert systems are the secure foundation of legal AI.
Neural networks are a breakthrough technology, but law requires strict adherence to the letter of the law.
In the context of the Russian legal system, expert systems remain the optimal choice.
They ensure control over each stage of the analysis, risk reduction and transparency for clients and government agencies.
While neural networks are learning to "understand" law, expert systems are already replacing routine, leaving lawyers with the most important thing — strategy and the human factor.
Hype around neural networks is often fueled by vendors of appropriate solutions. However, in the legal field, security and transparency remain priorities.
If you want to learn more about how to create an expert system or automate legal work, pay attention to the Botman.one low-code platform. With it, you can create a push-button chatbot for Telegram or VKontakte, automate business processes and develop algorithms without programming.
Subscribe to our Telegram group to keep up to date with news about #legaltech, #lowcode and automation of legal work: link.