I. Epistemology on the Differences Between Knowledge About the Future and the Past/Present

To explore the peculiarities of future cognition, this subsection draws mainly on insights from the branch of philosophy known as epistemology. Epistemology is the philosophical study of the nature, origin, and limits of human knowledge.Footnote ³⁰ Epistemological insights are widely used in legal scholarship, particularly in evidence law, which is also concerned with the acquisition of knowledge about the facts of legal cases. As a result, a distinct research field—legal epistemology—has emerged to examine the epistemological questions raised by legal practice.Footnote ³¹

As discussed, according to general evidence law, one of the main aims of a legal process—the trial—is to seek the truth. However, truth is the immediate and general aim. To the extent that truth serves as the standard of correctness for a verdict or finding, the ultimate aim may be said to be knowledge.Footnote ³² The same applies to future fact-finding: Its ultimate aim is knowledge about the future.

Philosophers have long discussed how knowledge of the past differs from knowledge of the future. They generally agree that we know more about the past than the future—a phenomenon often referred to as the temporal asymmetry of knowledge. In simplified terms, the laws of physics are time-symmetric: They allow us, in principle, to infer both past and future states from present data with equal certainty. For example, we might know that the sun rose yesterday and will rise tomorrow by applying the same physical laws. However, when it comes to the past, we have an additional source of knowledge: records, also referred to as signs, traces, memories, evidence, and so forth. While our belief that the sun will rise tomorrow is based only on inference from physical laws, our belief that it rose yesterday can also be based on direct experience or stored evidence—for instance, seeing the sunrise with our own eyes and remembering it.Footnote ³³

Accordingly, people usually acquire knowledge about the past and present and future in different ways. Philosophers have proposed various classifications of these methods. For instance, Stephan Torre distinguishes between downstream and non-downstream processes of acquiring knowledge. Downstream processes involve perception, memory, and testimony. Downstream processes are different from non-downstream processes because they all involve a causal chain from the event that one has knowledge about to one’s knowledge-constituting belief. Downstream processes are generally considered reliable, but they cannot be used to obtain knowledge about the future, as the subject is not causally downstream of those events. For example, if one sees footprints in the sand, it means that someone has left them there after walking on the sand in the past, but never in the future. Non-downstream processes include induction, mental simulation, and practical foreknowledge. These non-downstream processes can be further explained. Induction involves reasoning from a finite number of observed instances of an event-type A having property F, to unobserved instances of A also having F. For example, because we have records showing it has snowed in Lithuania for the past 100 winters, we can reasonably infer that it will likely snow again this winter. Mental simulation refers to our use of an “intuitive physics engine,” a cognitive model of how physical objects behave. This allows us to predict future states with substantial accuracy. For example, seeing a glass roll toward the edge of a bar, we form the belief that it will fall and shatter. Similarly, people also have “intuitive psychology engine,” which allows us to simulate the mental states of others and thereby their future mental states and actions. Practical foreknowledge is when a person knows what will happen because they intend to make it happen—for example, knowing one will go for a run later that day.Footnote ³⁴

Richard Swinburne, similarly to Torre, distinguishes between inferential and non-inferential knowledge. Non-inferential knowledge includes direct experience and traces, which cannot be obtained about the future, like downstream processes mentioned above. Inferential knowledge, like non-downstream processes, however, can be applied to future fact-finding.Footnote ³⁵ According to English dictionaries, inference is “a conclusion reached on the basis of evidence and reasoning”Footnote ³⁶ ; its antonyms include “fact,” “surety,” and “certainty.”Footnote ³⁷ Nicholas Rescher likewise notes that prediction is a matter of thought rather than perception. Where direct perception of any sort fails us, rational reflection based on available information can nevertheless penetrate.Footnote ³⁸ However, prediction has inherent limits—an idea that connects to David Hume’s classical critique of inductive reasoning. According to Hume, we have no actual warrant for an assumption that the future will resemble the past.Footnote ³⁹

That human knowledge about the future is more limited than about the past or present illustrates Torre’s assertion that in assessing whether a subject knows a proposition concerning the future, or whether assertions about the future are appropriate, people typically apply different considerations from those involved in assessing knowledge and assertions about the past. Consider a person who prepares a dish he has made a hundred times before, always to his partner’s liking. At the moment of serving, it is reasonable to say, “she will like this,” based on inductive reasoning and past experience. However, if asked the next day—without receiving any feedback—whether the partner liked the meal, it would seem less appropriate to reply, “she did.” Even though the underlying evidence is unchanged, past-directed assertions tend to require confirmation through downstream methods. For future knowledge, people accept non-downstream processes, while for past knowledge, they usually expect downstream confirmation. This contrast illustrates Torre’s point: Our standards for attributing knowledge differ depending on whether a proposition concerns the future or the past, and on the kind of belief-forming process involved.Footnote ⁴⁰

II. Recognition of Epistemological Limitations in Legal Practice

Legal practice recognizes the epistemic limits and specific features of future fact-finding—such as the fact that participants in legal proceedings cannot “prove” future events to the same extent as past or present facts, or that knowledge of the future and the past/present is acquired through different methods.

The European Court of Human Rights (ECtHR) case law in the so-called non-refoulement cases is an illustrative example. These concern whether a state’s decision to deport an applicant breaches Article 3 of the European Convention on Human Rights (Convention), which prohibits torture, due to the risk of the applicant facing ill-treatment—for example, torture—in the receiving country. Such cases are classic examples of future fact-finding in law, as they involve an assessment of a potential future event—whether the applicant would be at risk of ill-treatment if deported.

A particularly notable contribution is found in the concurring opinion of Judge Zupančič in the case of Saadi v. Italy. There, he conceptually explained the epistemic limits and evidentiary particularities of future fact-finding in a legal context. He observed that in such cases “judicial assessment does not have to do with a past historical event,” which is exceptional in legal practice because “the legal paradigm is retrospective,” along with “all its evidentiary apparatus.”Footnote ⁴¹ According to him, the “cognitive approach to future events may be only a rational probabilistic assessment in the spectrum of experiment which moves from abstract probability to concrete probability.”Footnote ⁴² While he acknowledged that fact-finding concerning past events also raises serious epistemic challenges, he stressed that “‘the evidentiary’ problem concerning future events is far more radical.”Footnote ⁴³ Zupančič emphasized that the correctness of probabilistic assessment—prognosis—“critically depends on the nature of information—not evidence!—adduced in a particular situation.”Footnote ⁴⁴ His distinction between information and evidence aligns with the epistemological distinctions previously discussed—such as downstream versus non-downstream knowledge, or non-inferential versus inferential knowledge, and between records—traces, signs, memories, and so forth—and natural laws. The majority judgment in Saadi v. Italy also acknowledged the epistemic limits of future fact-finding, noting that “assessment of that risk is to some degree speculative.”Footnote ⁴⁵

In other cases, the ECtHR has recognized these epistemic limits even more explicitly. In Rustamov v. Russia, the Court stated that: “[R]equesting an applicant to produce “indisputable” evidence of a risk of ill-treatment in the requesting country would be tantamount to asking him to prove the existence of a future event, which is impossible, and would place a clearly disproportionate burden on him.”Footnote ⁴⁶

In Azimov v. Russia, the Court similarly held that such an allegation:

In these non-refoulement cases, which are forward-looking, the ECtHR applies a different standard of proof than in “regular” backward-looking cases.Footnote ⁴⁸ The standard applied is whether there are “substantial grounds for believing that, if the measure complained of were to be implemented, he [the applicant] would be exposed to a real risk of being subjected to treatment contrary to Article 3.”Footnote ⁴⁹ In contrast, in backward-looking cases, the ECtHR requires that a breach of the Convention be proven “beyond a reasonable doubt.”Footnote ⁵⁰ Scholars have argued that the “substantial grounds for believing” standard is lower than the regular standard of “beyond a reasonable doubt.”Footnote ⁵¹ This difference is probably dictated by a common-sense understanding of the inherent limitations people face when trying to acquire knowledge about the future.

In merger review case law, the ECJ also similarly said that the prospective analysis of the effects of concentration on the market carried out by the Commission “are necessarily more uncertain than ex post analyses” under Articles 101–102 TFEU.Footnote ⁵² Based on this, the ECJ rejected the party’s argument that the Commission must “meet a particularly high standard of proof” in concentration cases.Footnote ⁵³ Put differently, the ECJ resisted raising the standard of proof too high, given the uncertainty which characterizes future assessments.

In conclusion, it is generally agreed, both in epistemological theory and legal practice, that human knowledge of the future, while possible to some extent, is more limited than knowledge of the past or present. There are more available sources of knowledge for the past and present, such as evidence and natural laws, than for the future, that is, natural laws alone, and people typically acquire knowledge about future events using different methods than those used for past or present ones.

I. Standard of Proof

Defining the standard of proof in a case is crucial because it determines the required level of certainty. In common law systems, civil cases generally apply the “balance of probabilities” or “preponderance of the evidence” standard, whereas criminal cases require proof “beyond a reasonable doubt” standard.Footnote ⁶⁹ In numerical terms, civil cases are often said to require a likelihood greater than 50%, while criminal cases require 90%.Footnote ⁷⁰

By contrast, the distinction between civil and criminal standards is less explicit in continental or civil law systems. Nevertheless, empirical research suggests that, in practice, their approach is similar to that of common law systems.Footnote ⁷¹

The difference in standards reflects the varying tolerance for error that society accepts in each legal context. This is often framed as “error-risk management.” The standard in criminal cases, on the one hand, is higher because wrongly convicting an innocent person is widely considered to be a graver mistake than erroneously acquitting a guilty person. A high standard of proof (for example, ninety percent) ensures that convictions of innocent people are relatively rare. On the other hand, in civil cases, the disutility of erroneously finding for or against the plaintiff is similar, which means the error-cost minimizing decision threshold is fifty percent. Therefore, in criminal law, the higher standard aims to avoid the more serious mistake of convicting an innocent person, while in civil law, the consequences of errors are generally seen as more balanced, justifying a lower threshold.Footnote ⁷²

The ECtHR illustrates this logic in its case law as well: To arrest a suspect, only “reasonable suspicion” is required, a level of evidence far below that needed for conviction or even the bringing of a charge.Footnote ⁷³ In error-risk terms, arresting an innocent person is considered less serious than convicting or indicting one. Accordingly, the definition of the standard of proof is also critical in future fact-finding. For example, in one of its judgments, the ECJ has considered which of the two alternative standards should be applied in merger review cases: whether a merger has “a strong probability” of significantly impeding effective competition or is “more likely than not” to do so. The ECJ chose the “more likely than not” standard because, in principle, the “strong probability” standard would be too high given the uncertainty of surrounding future evaluations.Footnote ⁷⁴ In light of this, this subsection explains the specificities of determining the applicable standard of proof in future fact-finding cases.

The first issue is whether the same standard of proof used in past or present fact-finding, such as “balance of probabilities” or “beyond a reasonable doubt,” applies to future fact-finding. These may be called “regular” standards of proof: the general evidentiary standards applied when courts assess whether a past or present fact has been established, respectively in civil and criminal cases. Two positions are possible. First, one could argue that it does. As in traditional fact-finding, where the aim is to establish truth, in future fact-finding the aim is to determine the likelihood of a future event. Therefore, the standard of proof remains the same; only the object of the inquiry changes—from truth to likelihood. For example, a similar position was taken by the UK Special Immigration Appeals Commission, which indicated in its decision that “we have asked ourselves whether the Secretary of State has satisfied us to a high civil balance of probabilities”—regular standard of proof—that the appellant is likely to continue to engage “in conduct that endangers the national security of the United Kingdom” if not deported.Footnote ⁷⁵ While the regular standard, whether “balance of probabilities” or “beyond a reasonable doubt,” helps the judge evaluate the evidence, the object of inquiry determines the level of risk—likelihood—necessary to apply a legal norm. For example, in Lithuanian insolvency proceedings, this separate standard for the court to reject the restructuring plan is a judge’s internal—subjective—conviction that “the proposed measures will obviously not help the legal entity overcome financial difficulties, preserve viability, or avoid bankruptcy.”Footnote ⁷⁶ Here, “obviously” is a central word because it indicates that the judge must be highly certain that the plan will fail to reject it. To paraphrase, in such a case, the court can reject the restructuring plan “if the judge is satisfied, on the balance of probabilities, that the plan will obviously fail.”

A second position is that the regular standard does not apply in future fact-finding. Instead, a unique standard is used, which coincides with the object of the future fact-finding, that is, the likelihood. To paraphrase the standard applicable in Lithuanian insolvency proceedings regarding a confirmation of the restructuring plan in this way, the court can reject the restructuring plan “if the judge is satisfied that the plan will obviously fail.” In other words, “balance of probabilities” no longer plays a role.

This Article supports the second position that in future fact-finding, a single special standard of proof is sufficient, and the regular standard does not have any role. This view is also more common in case law. For instance, ECtHR Judge Zupančič has argued that applying any traditional standard of proof to future events is incoherent because “one cannot prove a future event to any degree of probability because the law of evidence is a logical rather than a prophetic exercise.”Footnote ⁷⁷ Zupančič said that the “cognitive approach to future events may be only a rational probabilistic assessment in the spectrum of experiment which moves from abstract probability to concrete probability” and that the “correctness of this probabilistic assessment depends on the nature of information—not evidence—adduced in a particular situation.”Footnote ⁷⁸ The ECtHR applies a different standard in future-oriented cases, such as non-refoulement, using the formulation “substantial grounds for believing” that a person would face a real risk of treatment contrary to Article 3—significantly lower than the standard of “beyond a reasonable doubt” typically required in backward-looking cases involving past eventsFootnote ⁷⁹ and even lower than “balance of probabilities” (more likely than not).Footnote ⁸⁰

The UK Supreme Court also rejects the application of a regular civil standard of proof—“balance of probabilities”—in future assessments. The Court said that in such cases “the whole concept of a standard of proof is not particularly helpful” because, in principle, it “is a question of evaluation and judgment”Footnote ⁸¹ and that the “existence of a risk can therefore arise from evidence which is sufficient to establish a possibility but falls short of proof on a balance of probabilities.”Footnote ⁸² Like Judge Zupančič, the Court observed that the “reasoning process which is appropriate to evaluating a risk is different from the reasoning process which is appropriate to deciding the facts in issue in civil litigation.”Footnote ⁸³ The UK Court has also importantly noted that even past or present facts used for future assessments—for instance, previous contacts with terrorists to prove a threat to national security—“do not have to be proved to have occurred on a balance of probabilities.”Footnote ⁸⁴

In interim relief cases, courts such as the ECJFootnote ⁸⁵ and the Court of Appeal of LithuaniaFootnote ⁸⁶ also expressly apply lower standards than in main actions, reflecting the summary character of such proceedings, the provisional nature of such decisions and the limits of future knowledge. In a recent opinion, ECJ Advocate General Medina similarly repeated the ECJ’s consistent position that in merger cases, the Commission conducts forward-looking economic analyses, which “cannot be as certain as its ex post analyses–under Articles 101 and 102 TFEU.”Footnote ⁸⁷ These examples support the view that, usually, the epistemic limits of future cognition naturally push toward a lower and unique standard of proof.

Beyond the general principle that future fact-finding typically requires less certainty than regular standards like “balance of probabilities,” there are several additional reasons the second position—omitting a regular standard of proof—should prevail. First, using the regular standard of proof might be confusing and result in the application of an incorrect standard. For example, according to the UK Special Immigration Appeals Commission’s position that “the Secretary of State has to satisfy the Commission to a balance of probabilities that the appellant is likely to engage in conduct that endangers the national security,” it might appear that the standard is “more likely than not” that the appellant will threaten to the national security. But the UK Supreme Court rejects this position because the “existence of a risk can therefore arise from evidence which is sufficient to establish a possibility but falls short of proof on a balance of probabilities.”Footnote ⁸⁸ In other words, the “balance of probabilities” is not the standard in such cases. According to the Court, in principle, the applicable standard in such cases is flexible because the “evaluation of the risk then depends on such factors as the degree of risk, the possible methods of addressing the risk, and the gravity of the consequences if the risk eventuates.”Footnote ⁸⁹ Therefore, it would be erroneous to apply the “balance of probabilities” standard in such cases.

Second, using the regular standard of proof might not only be confusing but also “epistemically awkward” thus also confusing, and impractical. For instance, the Texas death-penalty statute requires prosecutors to prove beyond a reasonable doubt that there is a probability that the defendant will be dangerous in future—“definitely maybe” situation. Legal epistemologist Susan Haack describes this as “iterated epistemological operators.”Footnote ⁹⁰ However, demanding certainty about uncertainty seems epistemically awkward. United States legal scholar Kevin M. Clermont proposes collapsing such iterations into a single measure when trying to determine how an event is likely because the cumulative likelihoods contribute to predicting one proposition’s existence.Footnote ⁹¹ Essentially, distinguishing these two “epistemological operators” is not necessary in a practical sense because both perform the same function—they reflect the judge’s internal conviction about past, present, or potential future facts. For example, in Lithuanian civil procedure, the regular standard of proof is defined as a certain degree of sufficiency of the evidence and of the judge’s conviction, at which the fact-finder concludes that a disputed fact has been established or disproved.Footnote ⁹² The Supreme Court of Lithuania calls this the “reasonable belief standard,” which, in principle, requires the judge to be reasonably and personally—internally—convinced of the facts of the case.Footnote ⁹³ Similarly, the special standard in future fact-finding also reflects the judge’s subjective—personal, internal—belief about the likelihood of a future event, not an objective probability. For instance, when a restructuring plan is said to “obviously” fail to prevent bankruptcy,Footnote ⁹⁴ the word “obvious” implies a high degree of subjective certainty regarding the occurrence of the asserted future event but not an objective probability. Therefore, having two separate standards—regular and special—is unnecessary because, in any case, ultimately, the special “future-oriented” standard determines the judge’s decision, making the regular standard of proof, such as “balance of probabilities,” redundant. To conclude, this argument of “epistemic awkwardness” also supports the view that one substantive standard is sufficient for future fact-finding.

Once a special standard for future fact-finding is accepted, the next step is to determine what that standard is in a given case. This quest may involve one or two layers of interpretation: identifying the legal standard by classic methods of legal interpretation and, where relevant, adjusting it in light of the factual context.

First, the legal standard may be stated directly in the text of a legal norm—linguistic interpretation—or derived through other interpretative methods. To recall, this standard helps to manage the risk of error. How this risk is allocated is a matter of legislative lawmaking. According to Alex Stein, “[m]oral considerations that inform risk-allocating decisions belong to the domain of politics,”Footnote ⁹⁵ and there can be different reasons that lead to different levels of risk tolerance.Footnote ⁹⁶ For example, the law may allow a court to reject a restructuring plan if it is “obvious” that the proposed measures will not prevent bankruptcy.Footnote ⁹⁷ This term, “obvious,” clearly indicates a high standard.

In other cases, the standard is not explicit and must be interpreted using other methods. This must be done carefully, with attention to managing the risk of error, and meet certain requirements. First, relying on Ronald Dworkin, substantively, adjudicative procedures must tolerate a similar risk of error for similar types of error. For example, if both consumption of tobacco and drugs are prohibited, and non-compliance with this prohibition incurs liability of $1,000, the standard of proof for proving both violations must be the same. Second, risk-allocating procedures must be applied consistently so that no party is arbitrarily assigned a different risk level. For instance, prosecuting violations of the tobacco prohibition should occur using the same standard of proof, irrespective of whether the accused party is male or female.Footnote ⁹⁸ These considerations help clarify how interpretation guides the identification of standards for future fact-finding.

For instance, the text of the rule on the substantive requirements for granting interim relief before the ECJ is not informative. In principle, it states that the measures are granted if there are (i) circumstances giving rise to urgency and (ii) the pleas of fact and law establish a prima facie case for the interim measure applied for.Footnote ⁹⁹ However, the case law and the doctrine have clarified that the first standard, urgency, is high, while the second one, prima facie, is low.Footnote ¹⁰⁰ It is also widely accepted that the standard of proof in interim relief cases is lower than the standard applied in the main proceedings. For example, the ECJ requires sufficient persuasive evidence and assertions, which, however, need not be “convincing” in the sense that, at the interim relief stage, the judge hearing the application is required to come to a provisional and not a definitive view of the position.Footnote ¹⁰¹ A similar position is taken by the Court of Appeal of Lithuania.Footnote ¹⁰² The ECtHR judge Zupančič likewise observes that when the Court decides on the application for interim relief in the non-refoulement cases, all the evidentiary apparatus works differently because this decision does not serve an ultimate decision over the subject matter; it only freezes the situation so that the court may have the time and the opportunity to make justice prevail. Accordingly, in such situations, the aim of the decision-maker is not truth-finding—for example, whether the applicant would be ill-treated if expelled—but simply creating conditions for truth-finding by preventing irremediable consequences in case of irretrievable expulsion. Therefore, in these instances, the applicant must only produce “a shadow of doubt,” whereupon the burden of proof shifts to the country concerned. This is known in evidentiary doctrine as “bursting the bubble.”Footnote ¹⁰³

In another example from migration law, the Supreme Administrative Court of Lithuania ruled that the required likelihood of a national security threat for revoking a foreigner’s residence permit is higher for long-term residents than for short-term ones. Although the national law explicitly did not differentiate between long-term and short-term residents, the Court relied on EU law, namely, the Long-Term Residents Directive,Footnote ¹⁰⁴ which grants stronger protection against expulsion to long-term residents.Footnote ¹⁰⁵ This case is another example of the error-risk management function played by the standard of proof. Notably, it implies that erroneously revoking permits for foreigners with long-term resident status is socially more condemnable than revoking permits from short-term residents. The rationale is that deportation harms long-term residents more, given their deeper ties to the country.

Another example is the EU mergers control system, which uses a standard of “more likely than not.”Footnote ¹⁰⁶ This “symmetrical system” is explained by the fact that this regulatory environment, unlike some others that weigh certain interests, health, for instance, more than others, such as economic benefits, for instance, is not based on a “precautionary principle,” and the benefits of mergers are explicitly recognized by the EU legislature. Therefore, the symmetric system of EU merger review tries to evenly distribute error costs.Footnote ¹⁰⁷ This position also implies that other regulatory fields, which are based on the precautionary principle, such as health,Footnote ¹⁰⁸ environment,Footnote ¹⁰⁹ or national security,Footnote ¹¹⁰ can distribute the risk of error unevenly, prioritizing the reduction of respective risks instead of, for instance, economic benefits or other rights—interests—of an individual.

Second, as mentioned, the applicable legal standard may also be shaped by another layer of case-specific factual considerations—factual standard. In certain scenarios, the party carrying the burden of proving future circumstances may face procedural or natural—epistemic—difficulties in producing information necessary for the future assessment. Such a party is disadvantaged because it carries a risk of non-persuasion, which means that if it does not adduce sufficient evidence, it loses the case.Footnote ¹¹¹ According to Judge Zupančič, such distribution of the burden of proof works very well with past events, but it does not work very well in a specific category of future fact-finding cases, such as in pre-trial arrest cases or interim relief cases at the ECtHR. In these instances, which Zupančič calls “emergency situations,” for example, when the applicant is arrested at an airport in order to be expelled, it would be “clearly absurd” to put the regular burden of proof on the applicant. This stems from both epistemic limitations of future fact-finding and procedural disadvantage of the applicant, who has limited access to evidence due to his arrest. Thus, in such a case, the burden and the risk of non-persuasion should be redistributed by placing a large portion of it on the expelling State.Footnote ¹¹²

The ECJ also acknowledges these challenges in its interim relief jurisprudence, that is, recognizes these procedural limitations to some extent and relaxes the evidentiary standard in specific cases. For instance, where, due to the urgency of the situation, the judge grants relief before hearing the other party, the evidential requirements are much less than when the judge is considering the matter in the light of the other party’s observations; at a later stage in the interim relief proceedings, the judge must inevitably be more skeptical.Footnote ¹¹³ In another case, the Court has acknowledged the lack of “convincing evidence” in support of the applicants’ contention but, noting the scale of urgency and the asserted harm, liquidation, suspended the demand of the payment so that the applicants would have additional time to make a further interlocutory application with additional evidence.Footnote ¹¹⁴

Sometimes, it is epistemic limitations—rather than procedural barriers—that prevent the party bearing the burden of proof from obtaining reliable evidence for future assessments. This issue is not unique to future fact-finding; similar difficulties can arise in past or present fact-finding as well. For example, in certain types of cases, such as bid rigging in public procurementFootnote ¹¹⁵ or unfair competition,Footnote ¹¹⁶ the Supreme Court of Lithuania explicitly lowers the standard of proof due to the structural epistemic challenges these cases present—namely, that in such cases, only circumstantial—indirect—evidence is typically available. These considerations are highly relevant to future fact-finding, which is also characterized by inherent epistemic uncertainty. When deciding such cases, the decision-maker must take into account the specific nature of the case and, even before engaging with the evidentiary material, must ask in abstracto: To what degree of certainty can the party carrying the burden of proof establish the likelihood of the asserted future event? What kind of evidentiary material is, in theory, accessible to that party? If the nature of the case suggests that the future assessment is highly speculative, the decision-maker must factor this into their approach and avoid imposing an unrealistic evidentiary burden on the party. As the legal maxim states, the law does not compel the impossible (lex non cogit ad impossibilia).

Another factual consideration which might affect the standard of future evaluation is the circumstances of a particular case, especially the risks involved. Here, error-risk allocation plays an important role, and the final standard depends not only on the applicable legal norm but also on particular circumstances of the case. For example, as already discussed, in a case regarding a deprivation of citizenship due to national security risk, the UK Supreme Court has described that the evaluation of such a risk depends not on a fixed standard, such as a balance of probabilities or other, but on such factors as “the degree of risk, the possible methods of addressing the risk, and the gravity of the consequences if the risk eventuates.”Footnote ¹¹⁷ As an illustration, the Court referred to terrorist attacks in Europe in recent years, such as the Bataclan attack in Paris and the Manchester Arena bombing, and observed that “a precautionary approach is necessary in the interests of public safety” because an “error in judgement could have catastrophic consequences.”Footnote ¹¹⁸ In other words, the Court has implicitly lowered the standard of the future assessment for depriving of citizenship in this case because an erroneous judgment not to revoke citizenship poses a more serious threat to society—terrorism—in comparison to an erroneous deprivation of a single person’s citizenship. It is another example of how error-risk management plays a role in future assessments by affecting the applicable standard. In another case, the House of Lords had explained that when deciding such cases, “it is necessary to take into account not only the degree of probability of prejudice to national security” and “the importance of the security interest at stake” but also “the serious consequences of deportation for the deportee.”Footnote ¹¹⁹ That is, when adjusting the standard, not only public interests—national security—must be considered, but also the impact of the decision on the person involved.

In conclusion, accurately defining the standard of proof in future fact-finding cases is essential, as it directly influences the outcome by serving as a tool for managing the risk of error. Owing to the inherent epistemic limitations of predicting future events, the applicable standard in such cases is distinct and more variable than in past or present fact-finding, but usually lower. It may be derived directly from the text of legal norms or established indirectly through interpretative methods. Additionally, this standard may need to be adjusted based on the specific circumstances of the case, particularly where procedural or epistemic constraints limit the party’s ability to present reliable evidence or other important factual considerations involved, such as the nature of risks at issue.

II. Reasoning and Decision-Making

After determining the purpose of future fact-finding and the level of risk necessary to support a finding, that is, the standard of proof, the next step is to consider how the fact-finder should arrive at that conviction. Even legal norms that explicitly require future fact-finding emphasize “reasoning” as a central element of the assessment. For example, grounds for arrest must be based on a reasonable belief that the suspect will flee from the authorities, obstruct the proceedings, or commit further crimes.Footnote ¹²⁰ Similarly, a regulator may refuse to authorize the acquisition of a credit institution if there are reasonable grounds to suspect that the transaction may increase the risk of money laundering or terrorist financing.Footnote ¹²¹ In non-refoulement cases before the ECtHR, the standard is phrased as “substantial grounds for believing” that the applicant would face a real risk of ill-treatment.Footnote ¹²² The UK Supreme Court has also observed that, despite inherent uncertainty, an assessment of future risk must have a basis in “objective evidence.”Footnote ¹²³ These formulations indicate that reasoning is essential in future assessments.

This aspect of reasoning and decision-making aligns with Acharya’s “rationality” requirement for a factually reliable fact-finding process. According to this principle, justice in truth-seeking depends not only on the accuracy of the outcome but also on the rationality of the deliberation that led to it. If the process is irrational, then—even if the final verdict is correct—an unjustifiable error has occurred. For instance, it would be wrong to convict a defendant based on a coin toss, even if the verdict happens to be correct and even if the evidence admitted in court supports the verdict.Footnote ¹²⁴ Therefore, for future fact-finding to be considered “factually reliable,” it must also satisfy the requirement of rationality. This subsection further explores these rationality requirements by drawing on previously discussed philosophical insights.

As discussed earlier, in past or present fact-finding, the evidentiary material available to the fact-finder includes both evidence, that is, traces or records of past events, and laws of nature. For instance, in a murder trial, the fact-finder may rely on both evidence—such as witness testimony—and laws of nature—such as the time of day and expected daylight hours—to conclude that it was likely bright at the time of the event. In future fact-finding, however, the fact-finder does not have access to traces of future events and must rely solely on laws of nature or similarly stable regularities. As ECtHR Judge Zupančič noted, the “correctness of that probabilistic assessment [ … ] critically depends on the nature of information—not evidence—adduced in a particular situation.”Footnote ¹²⁵ This means that the accuracy of future-oriented findings depends entirely on the reliability of the reasoning methods used, as evidence about future circumstances does not exist.

From a general theoretical standpoint, philosopher Rescher argues that rational prediction is only possible when the future is foreshadowed in the stable and observable patterns of the past and present. He distinguishes between mere speculation and rational prediction, asserting that only a “rational prediction that has a basis whose merits are discernible prior to the event is of serious cognitive interest.”Footnote ¹²⁶ This distinction is essential in law to differentiate “reasonable” from “unreasonable” future assessments. According to Rescher, any sort of rational prediction requires informative input material satisfying three conditions: (i) data availability—the ability to obtain timely, accurate, and reliable information about past and present conditions; (ii) pattern discernibility—the existence of observable patterns in this data; and (3) pattern stability—the expectation that these patterns will persist into the future. Rescher says that rational prediction pivots on the existence of some sort of appropriate linkage that connects our predictive claims with the input data that provides justification. This linkage can be based either on explicitly articulated principles, such as explanatory regularities or presumed laws of nature—so-called formalized/inferential/discursive prediction—or on personal judgments that exploit a knowledgeable expert’s tacit, unarticulated, and sometimes inarticulable background knowledge about this matter—unformalized/judgmental/intuitive prediction. Thus, according to Rescher, we can have justified confidence in a prediction if (i) we are confident in the accuracy of the data and either (ii.a) the validity of the inferential principles being applied or (ii.b) the competence of the expert making the judgment. In either case, what matters is whether the predictive method has a reliable track record. Yet, Rescher also notes that these general principles do not resolve all challenges of prediction, because different domains, for instance, meteorology, economics, and astronomy follow different methodologies.Footnote ¹²⁷ The same is true in legal reasoning, which must be context sensitive.

A practical illustration of these theoretical insights is found in the ECtHR’s case law on non-refoulement. These cases involve a relatively consistent forward-looking methodology for assessing whether the applicant would face a risk of ill-treatment if deported. Various concepts already discussed—such as Torre’s non-downstream processes of acquiring knowledgeFootnote ¹²⁸ and Rescher’s emphasis on data and linkage—are evident in this jurisprudence. To determine the existence of a risk, the ECtHR examines the foreseeable consequences of deportation by analyzing both the general situation in the receiving country—objective test—and the applicant’s personal circumstances—subjective test.Footnote ¹²⁹ For the general situation, the Court relies on reports from international organizations, domestic authorities, and NGOs.Footnote ¹³⁰ In Rescher’s terms, such sources provide the “input data” necessary to identify and confirm patterns—such as systemic ill-treatment of specific groups in a given country. However, ECtHR judges do not always agree on how to interpret such data. For example, in Chahal v. the United Kingdom, the judges disagreed on whether the applicant’s “high profile” in the Sikh separatist movement in India lowers or raises this risk.Footnote ¹³¹ In such situations, the disagreement must be resolved by referring to past examples—for instance, whether the receiving country, India, has a track record of mistreating individuals in similar positions. The ECtHR also considers diplomatic assurances from the receiving state, that is, obtained assurances or guarantees from states to which a person is to be extradited or expelled, to the effect that that person will be treated in accordance with fundamental human rights.Footnote ¹³² From Torre’s perspective, such assurances resemble “practical foreknowledge,” as they reflect the intentions of the state providing them. Lastly, in the non-refoulement cases, the Court evaluates the totality of the evidence using a kind of “mental simulation”—imagining potential scenarios that might unfold if the applicant were deported. This form of reasoning involves projecting likely outcomes based on all available data. This short overview of the ECtHR case law indicates that the Court recognizes the methodological specifics of future fact-finding, at least to some extent.

Merger review also offers a clear example of how said principles contribute to future-oriented legal fact-finding. According to the EU Horizontal Merger Guidelines, very large market shares—typically around 50% or more—may themselves indicate a dominant market position.Footnote ¹³³ This reflects inductive reasoning, as past experience shows that such levels of concentration often lead to significant impediments to effective competition. Authorities may also rely on evidence that “the merging parties intend or expect the merger to lessen competition,”Footnote ¹³⁴ which exemplifies practical foreknowledge. “Merger simulation”—a technique that combines explicit theoretical and empirical approaches to predict the price effects of a mergerFootnote ¹³⁵ —can be seen as, in Rescher’s term, a formalized prediction, as opposed to an unformalized prediction, which rests upon an expert’s tacit and unarticulated knowledge. An example of such an unformalized prediction in law could be a situation when a financial markets regulator assesses a business plan of a company for licensing purposes in order to check the viability and sustainability of the business model.Footnote ¹³⁶ This assessment is mostly based on the unarticulated expertise of the regulator, rather than explicit principles.

In conclusion, future fact-finding requires a fundamentally different reasoning process than that used for assessing past or present facts. Because there is no direct evidence of future events, the fact-finder must rely entirely on inferential—non-downstream—methods. Legal formulations such as “reasonable belief” or “substantial grounds for believing” reflect the importance of well-reasoned, structured assessment rather than speculative guesses. The fact-finder must base conclusions on reliable present or past data, observable and stable patterns, and a credible link between known facts and projected outcomes. The fact-finder should be methodologically aware and combine structured input with necessary judgmental principles to make rational, justified predictions. Ultimately, future-oriented reasoning is not about proving certainty, but about constructing a justified, rational pathway from what is known to what is likely.

III. Ex-Post Review

How should past assessments about the future be evaluated when, by the time of review, the future is already known? For example, how should a court assess the legality of government-imposed COVID-19 measures that severely restricted individual rights but later proved ineffective in controlling the virus? To illustrate, the ECJ has examined the legality of Belgian travel bans during the COVID-19 pandemic,Footnote ¹³⁷ although the effectiveness of similar measures in combating the pandemic later was questioned.Footnote ¹³⁸ Should the court simply conclude that the measures were unlawful because they ultimately failed, or should the assessment be more nuanced?

This issue aligns with one of the Acharya’s requirements of a factually reliable fact-finding process—the admissibility of evidence. Can information about the actual accuracy of a future assessment be admissible for ex post legal evaluation of the assessment? For example, now, there is a pending case before a Lithuanian court, where the applicants—airlines—contest airport charges because, among other things, they were based on inaccurate traffic forecasts.Footnote ¹³⁹ However, the accuracy of these forecasts is already known, even though the case is not finished yet. Should the court evaluate this accuracy, or only stick to the information available to the forecaster at the moment the forecast was made?

Political psychologist Philip E. Tetlock, whose work includes one of the most comprehensive empirical studies on forecasting, and journalist Dan Gardner, who helped present this research in a book, argue that the right question in such a review is not whether the forecast was correct, but whether it was reasonable. Answering this question, they say, requires a form of “mental time travel”—placing ourselves in the decision-maker’s position at the time of the decision, with access only to the information that was available then. According to them, it is not oxymoronic to say that a prediction can be both wrong and reasonable. However, in their opinion, such a “mental time travel” is psychologically difficult because a situation as that tempts people with bait and switch: Replace the tough question—whether it was reasonable—with the easy one—whether it was wrong—answer it, and then sincerely believe to have answered the tough question.Footnote ¹⁴⁰ The same reasoning applies to assessing forecasts of risk: If a meteorologist predicts a seventy percent chance of rain and it does not rain, that forecast is not necessarily wrong—it still reflected a thirty percent chance that it would remain dry. A single outcome does not invalidate a probabilistic judgment. The only way to know for sure would be to rerun the day hundreds of times—if it rained 70% of those times and did not in thirty percent, the forecaster was correct.Footnote ¹⁴¹

It seems that this approach of an ex-post review of future assessments based on the criterion of reasonableness prevails in a legal context as well. For instance, in the non-refoulement case Vilvarajah and Others v. UK, the ECtHR concluded that the UK government did not breach Article 3 of the Convention by expelling the applicants to Sri Lanka, even though they were ill-treated after the expulsion. The Court relied, inter alia, on the principle established in its case law that the existence of the risk must be assessed primarily concerning those facts which were known or ought to have been known to the Contracting State at the time of the expulsion. For this case, the Court said that even though the applicants were ill-treated, there existed no special distinguished features in their cases that could or ought to have enabled the UK to foresee that they would be treated in this way.Footnote ¹⁴²

In merger review cases, the ECJ has a consistent view that the legality of the contested decision of the Commission regarding the merger must be assessed based on the facts and law existing at the date on which the decision was adopted and the information available to the Commission at that time; an applicant cannot therefore rely before the ECJ on facts subsequent to the decision.Footnote ¹⁴³

A similar reasoning framework appears in Lithuanian case law concerning directors’ liability under the business judgment rule. The courts apply this rule when deciding whether a company’s director is liable for a business decision which was financially harmful to the company, for example, a failed investment. This rule protects bona fide directors from damage claims in order to ensure that directors can take reasonable risks, which is essential for business development. According to the rule, the mere fact that the director’s business decision had negative economic consequences for the company is not sufficient to hold the director liable. He would be liable only if he acted in bad faith, breached legal imperatives, was disloyal to the company, or acted recklessly—for example, took an unreasonable risk. The former, recklessness, is important in the context of future fact-finding. When courts investigate whether the director had acted recklessly, they check, for instance, whether he got acquainted with all the necessary information before deciding, whether he evaluated whether the risk was reasonable considering the economic benefit pursued, took into account the company’s financial standing and creditors’ interests, and so forth.Footnote ¹⁴⁴ In other words, courts do not impose a legal obligation on the directors to foresee the future when making business decisions. Directors may fail and are not liable for that as long as their decisions remain reasonable in light of available information at the time of the decision.

In conclusion, returning to the example raised at the start of this subsection: No, a court should not find that COVID-19 control measures were illegal solely because they later proved ineffective. The government does not have a divine power of foreknowledge. The appropriate legal test is reasonableness—whether, based on the information reasonably available at the time, it was reasonable to believe the measures could be effective. If the answer is yes, then the measures should be considered lawful. Information that was unknown or unknowable at the time should not be admissible as evidence in an ex post legal evaluation. Anything else would conflict with the principle lex non cogit ad impossibilia. In principle, the ECJ followed this approach in its review of the Belgian travel bans. The Court emphasized the precautionary principle, according to which a Member State should take protective measures without having to wait until the reality of those health risks becomes fully apparent. It also stated that national courts must assess the appropriateness of such measures “in the light of the scientific data commonly accepted” at the time of adoption and “in view of the degree of uncertainty that might prevail.”Footnote ¹⁴⁵