There will be a major upheaval in the credit industry at the end of this year. Libor (London Interbank Offered Rate), a reference interest rate for loans, will be abolished, and contracts with this interest rate will have to be rewritten. This will affect contracts worth around $350 trillion and hundreds of thousands of bank customers. Now, many customers will not know about this, or financial institutions could let Libro’s sales remain in the contract because of shoddy work or intent to defraud.
Now the number of contracts to be sifted through is enormous, and courts can fill weeks and months of sitting days just with these cases. Here, however, the technology of the decade is applicable to accomplish this task: Artificial Intelligence.
Here, an algorithm can scan all the digital contracts through text recognition and work out Libro. The AI understands even the context, what the deal is about, and why the word Libro appears. Based on this knowledge, a decision can be made regarding whether the contract is illegal or not. Furthermore, it is also possible for the algorithm to rewrite the contracts, replacing the Libro with a valid reference interest rate. This saves enormous resources at the institutions and courts and provides an enormously high degree of certainty that new laws are applied correctly.
Now that an AI is already making judgments and rewriting contracts, the question is:
When will I be represented in court by a robot?
This is not likely to happen soon. An AI needs useful labeled data that describes the context to get a good impression of the situation. In the case of contracts, this is not a problem because the context becomes clear based on the entire agreement, and the AI can work out a reasonable solution.
In a negotiation, this is not always a given. An AI can process evidence, mandates, and pleadings; even incorporating the law is not a problem. However, many semantic elements are needed in the negotiation that an AI can still misinterpret. Decision and interpretation margins, agreements based on certain developments, and evaluation of the given liability clauses are difficult. An AI basically understands a yes or no, a 1 or a 0. Intermediate values are feasible, but this makes the algorithm more complex, and the AI needs more training data.
AI will, therefore, be able to play a significant part in adjudication and save time and enable further improvements. The AI can fight discrimination and corruption because the evaluation is sober and based on concrete facts. However, AI will only be able to do this when evaluating documents/evidence and making suggestions for the jurisdiction. The AI will also have to deal with the legacy of human jurisprudence. If the AI is trained with existing judgments and various discriminations are present here, the AI will initially also work in a discriminatory manner. This must be invalidated based on non-discriminatory elements, such as suitable judgments or the laws‘ comparison. Furthermore, evaluating the AI and supervised learning about the judgments made is a proven method to obtain a neutral AI.
Nevertheless, the use of machines in the administration of justice shortly is not utopian. Many courts and lawyers are calling for permission to use machine evidence. For example, if an assistance system warns of imminent danger while driving or at construction sites and an accident occurs because the person ignored the warning, it must be possible to determine fault based on the machine data. The next step should be the use of machines in the courtroom without a decision-making function, for example, as a lie detector during witness testimony or examination of evidence, among other things, in cases of suspected falsification of evidence.