I. Mitigation by courts following-Kashmira Singh: An adherence to the Bayesian Framework

The Bayesian framework relies on the likelihood ratio. This is a measure of the strength of the new evidence. In the earlier scenario, the likelihood ratio is less than one because the new/circumstantial evidence gave the expected result only because of the corroborating evidence of the confession. Since this ratio is less than one, ideally limited reliance should have been placed on it and that too, independently. In other words, it should not have been used to strengthen an already-weak piece of evidence. 

This correct, Bayesian approach has been laid down in Kashmira Singh where the Supreme Court held that the proper way to approach a confession by a co-accused is, first, to marshal the evidence against the accused excluding the confession altogether from consideration and see whether, if it is believed, a conviction could safely be based on it. If it is capable of belief independently of the confession, then of course, it is not necessary to call the confession in aid. But cases may arise where the judge is not prepared to act on the other evidence as it stands even though, if believed it would be sufficient to sustain a conviction. In such an event, the judge may call in aid the confession and use it to lend assurance to the other evidence and thus, fortify himself in believing what without the aid of the confession he would not be prepared to accept. This rationale of the court mitigates both the issues that arise from the literal interpretation of Section 30.

Firstly, the issue of system inefficiency arising from the imbalance between the error costs and the costs of avoiding the error are mitigated as Kashmira Singh makes the confession corroborative evidence, which can be used in an extremely limited sense. In doing so, it adds the safeguard of corroboration to tilt the balance in favour of the accused as it ensures that the court does not abdicate its duty of fact-finding by relying only on the confession. This is evident from the fact that courts have placed almost zero reliance on confessions since it does not fit the Kashmira Singh rationale. In Haricharan Kurmi v. State of Bihar, the Court reiterated that confession of a co-accused cannot be the primary evidence and only can lend reassurance to reaffirm or counter the inferred conclusion. Even in cases involving special legislations, courts have read-in the corroboration element into the testimony of a co-accused. In State (NCT of Delhi) v. Navjot Sandhu, the question was concerning the interpretation of Section 15 of TADA, which made the confession by a co-accused admissible in the designated court. The mandatory nature of the provision deviated from the discretion granted under Section 30. Despite this, the court held that section 15 does not necessarily imply a departure from the normal rule of corroboration as is usually followed in cases under Section 30, even if the section does not obstruct such confessions from forming the sole basis for conviction.

Secondly, the issue of confirmation bias is also resolved by Kashmira Singh’s reasoning as it adheres to Bayes’ Rule. It makes confessions admissible only in those cases where the other evidence available is ordinarily sufficient to base a conviction. In doing so, it solves the earlier problem wherein confessions were used to lend credibility to other, inherently weak evidence which was then used to justify convictions. This allows the court to follow a strict separation between prior beliefs and the likelihood of other evidence conforming to these beliefs. 

A Bayesian considers both these pieces of evidence independently instead, of viewing them on the basis of the order in which they are presented. This allows the Bayesian judge to avoid double-dipping of data wherein the judge uses the evidence first obtained to update her beliefs about the credibility of circumstantial evidence and accordingly to formulate her likelihood ratio. A Bayesian formulates her prior beliefs but considers them to be independent of the likelihood of the data yielding a particular outcome or conclusion. This can be seen in Kashmira Singh, as it prescribes a two-step process wherein the court is to consider the evidence available, excluding the confession, independently. It is only if the judge is unsure in her belief in the available evidence (though objectively, the existing evidence is sufficient to convict), can the judge minimally rely on the confession. In most cases, either the available evidence is extremely weak/unreliable or it is strong enough to sustain a conviction. This further explains why post-Kashmira Singh, courts have hardly resorted to the confession by the co-accused. 

The next section of the paper analyses the change in this position by the BSA.

II. BSA and the chipping away of a safeguard

Section 24 of the BSA is the equivalent of Section 30 of the IEA. But there is a significant difference between the two. Explanation II to Section 24 creates a deeming fiction where trial of an absconding accused shall be deemed to be a joint trial for the purposes of this section. This in effect, chips away the safeguard of joint trial which had been embedded in Section 30. While it remains to be seen as to how courts would interpret Section 24, there is a strong chance of reversion to the pre-Kashmira Singh understanding of the admissibility of the confession of the co-accused. 

This is because the deeming fiction has a considerable implication as it allows the court to proceed against an absconding accused by relying on the confession of a co-accused. Since the physical and mental presence of an accused is not required for conducting a joint trial, in the absence of any other cogent evidence owing to the abscondence, the only evidence that may be available with the court may be the confession of a co-accused under Section 24. This would again bring up the same problems of inefficiency and confirmation bias as it would significantly dilute the corroborative evidence holding of Kashmira Singh. This would result in increased error costs and allow courts to yet again rely on the confession to lend credibility to weak evidence. 

Conclusion 

While courts have been proactive in ensuring the minimal use of the confession of a co-accused under Section 30, it remains to be seen as to how Section 24 changes the position. The omission of “joint trial” is concerning as it ties the hands of courts in terms of attributing a narrow construction to Section 24 of the BSA. This could have a detrimental impact on the rights of the accused.

Aakriti Rikhi is a law student at National Law School of India University, Bangalore.