The specialty of Cross-Examination assumes an essential job in the preliminary of each case which includes diligent work and ability of attorneys while giving equity to their customers. While manual cross-examination may not be altogether done away with, we shall see if it can be supplemented with AI to the advantage of the advocates and reduce skill-dependency of individuals. This is turn may help clients through reduced cost of litigation; aside from the increased legal efficiency.
What Cross-Examination Is
The examination of an observer by the antagonistic party will be called his cross-examination. The motivation behind the interrogation is to test the veracity of the observer. The fundamental object of the act is to discover reality and identification of misrepresentation in human declaration. It is structured either to demolish or debilitate the power of proof which is effectively given by a chance observer. Questioning of observer is an obligation of each legal counsellor towards his customer and not a matter of wonder and notoriety. It is perhaps the strongest test to find reality and to identify the bogus articulations of the witness. It ought to be recalled that the Justice ought not be miscarried by the inappropriate or inadequate cross-Examination.
Application of AI for Cross-examination
Possible Application
If we remove the subjective components of cross-examination, which can vary from advocate to advocate, we can come across a set of processes which can be driven algorithmically. Of course, it would have to depend on sensors, voice recognition and cognitive algorithm as substitute for the senses of the human lawyer. But it would not be impossible to drive a decision system based on the logical line of questioning if the facts of the case are fed to it beforehand. This of course, would not be sufficient in itself. But it can serve as a first line of examination or a cross-check to the efforts of the human lawyer; with an objective to become a viable supplement as technology advances. There can be cognitive logic to figure out where the witness may be lying e.g. increase in pulse rate, blood pressure, inability to look to the right, non-movement of any other body part apart from mouth at the time of talking and blinking of the eye too frequently. The subsequent questions can be framed based on these determinations of the AI machine; the flow of questions would be derived based on the general logic of questioning; but customized to what it encounters are the other end to divert the flow of the program, based on the intelligence logic.
With sufficient computing power, the three-pronged inputs i.e. case details in the form of facts in issues and relevant facts – with their mapping for a particular question, the answer/s to the preceding question/s and the feedback in the form of sensor inputs; can be together compiled and interpreted at run-time sufficiently fast; so that as to dynamically come up with the next question which would be meaningful. These will, together form the series of questionnaire for cross-examination of the witness. But it has to take into account the different types of witness, their credibility and reliability and the types of cross-examination based on that.
Challenges Due to Vast Array of Witness Categories
The broad spectrum of witness types mentioned earlier make the task of bringing in automation in the field of cross-examination considerably more difficult than in other areas of law such as citation search or writing standard form contracts. It is to be kept in mind that the basic purpose of software is to do large volume of repetitive work. While AI has definitely pushed this bound, it is very challenging to handle the entire dynamics of cross-examination as mentioned above.
Still possibility exists and perhaps instead of a single AI, there can be different modules of AI specializing in handling the different types of witness and case situations. The algorithms and the big data datasets from past cases would vary in each case and each AI would thus specialize in handling specific cases and would be having the artificial intelligence by imbibing data for specific types of witnesses.
For example, the AI enabled to cross-examine ordinary witnesses may be based on the following premises laid down by the Supreme Court and its internal logic and machine learning would be calibrated based on these. “The premises are: By and large a witness cannot be expected to possess a photographic memory and to recall the details of an incident. It is not as if a video tape is replayed on the mental screen.
Ordinarily it so happens that a witness is overtaken by events. The witness could not have anticipated the occurrence which so after has a statement of surprise. The mental faculties therefore, cannot be expected to be attuned to absorb the details.
The powers of observation differ from person to person. What one may notice, another may not. An object or movement might emboss its image on one persons mind whereas it might go unnoticed on the part of another.
By and large people cannot accurately recall a conversation and reproduce the very words use by them or heard by them. They can only recall the main purport of the conversation. It is unrealistic to expect a witness to be a human tape recorder.
In regard to exact time of an incident, or the time duration of an occurrence usually, people make their estimates by guess work on spur of the moment at the time of interrogation and one cannot expect people to make very precise or reliable estimates in such matters. Again it depends on the time sense individuals which vary from person to person.
Ordinarily a witness cannot be expected to recall accurately the sequence of events which take place in rapid succession or in a short time span. A witness is liable to get confused or mixed up when interrogated later on.
A witness though wholly truthful, is liable to be overawed by the court atmosphere and piercing cross -examination made by counsel and out of nervousness mixes up facts gets confused regarding sequence of events. Or fills up details from image on the spur of moment. The subconscious mind of the witness sometimes so operates on account of the fear of looking foolish or being disbelieved though the witness is giving a truthful and honest account of the occurrence witness by him perhaps it is a sort of psychological defense mechanism activated on the spur of the moment.
People react to situations not always in a uniform way. An educated, an illiterate a city dweller, a villager or an adivasi will react differently according to the degree of their sophistication. Moreover even in the case of the same class, the reaction would vary with the physical courage, mental equipment and social awareness of the individual” .
For other category of witnesses, the cross-examining AI would need to be trained and its internal logic would be different. 
AI still has a long way to go in being applicable to interactive situations like cross-examination. But the basic logic, the hardware and the algorithms are there to enable a start or at least take a shot to create a prototype. It would be governed by a combination of the facts of the case, the answers of the witness and the feedback from the sensory receptors like camera and monitors. The admissibility of the answers would not be questionable unlike lie-detector or narco-analysis, which are not proven as they interfere directly with the state of the witness. AI on the other hand will only be questioning; not interpreting answers to the court. So the admissibility of the cross-examination as evidence would not be an issue.


  • Maganlal v King Emperor AIR 1946 Nagpur 126
  • Juwar Singh v State of M.P, AIR 1981 S.C 373
  • Md. Ibrahim Khan v Susheel Kumar And Anr, AIR 1983 AP 69
  • Evidence Act 1995 (Cth) s 41(1); Evidence Act 1995 (NSW) s 41(1) – 
  • https://www.alrc.gov.au/publications/28.%20Other%20Trial%20Processes/cross-examination#_ftn167
  • https://www.law.cornell.edu/rules/fre/rule_611
  • Repository.binus.ac.id
  • Repository.binus.ac.id
  • Bryan a Garner (Ed.), Black’s Law Dictionary, p.1596.(West group, St. Paul, Minnesota, 17thEdn., 1999)
  • Halsbury’s Laws of India
  • Chapter IX titled “OF WITNESSES” of the Indian Evidence Act, 1872
  • Historic Doubts relative to Nepoleon Bona Parte, P.14 Sixth Ed., C.D. Fields’ Law of Evidence in 11thEd. Vol. 1 P. 19.
  • Vadivelu Thevar v State of Madras, AIR 1957 SC 614
  • State of Punjab v. Tarlok Singh, AIR 1971 SC 1221
  • 1957 Cri LJ 32
  • Koli Nana Bhana v State of Gujrat, 1986 Cri. L. J. 571
  • The State of Punjab v Hari Singh, A.I.R. 1974 S.C. 1168
  • 1983 Cri.L.J. 1096 (S.C.)
  • Bhagwan Singh v State of Bihar – AIR 1976 SC 202
  • Bhogin Bhat Kirji v State of Gujarat,1983 Cri.L.J. 1096 (S.C.)

Leave a Comment

Your email address will not be published. Required fields are marked *