Proving research misconduct

Abstract

Detecting and punishing violations of research integrity requires first having to prove them. However, establishing proof of research misconduct presents a number of challenges. Firstly, it has to be conducted in a variety of contexts, including before research integrity officers, university disciplinary committees, civil courts, criminal courts, at first instance and on appeal. In each of these instances, the rules of evidence have their own specific features. Secondly, it may be necessary to prove not only material facts, but also circumstances that are more difficult to grasp, such as intention or negligence. In addition, not all evidence is admissible, as it may breach a duty of fairness or protected secrets. Consequently, research integrity officers and judges may be unable to consider all relevant evidence when assessing allegations of misconduct. This article examines the legal issues pertaining to the proof of research misconduct, based on French law and the law of the European Convention on Human Rights. This article examines the burden of proof, the object of proof, the admissibility of evidence and the exclusion of evidence. It identifies the differences in evidence that exist depending on whether investigations into research misconduct are conducted before research integrity officers or before the courts. It analyses the justifications for these differences and concludes that the role of research integrity officers should not be confused with that of judges, even if this leads to differences in the proving of research misconduct.

Proving scientific fraud is at the heart of the effort to curb research misconduct and constitutes a pivotal element in the promotion of research integrity. Solidly establishing the existence of misconduct is an essential first step before making it public and imposing sanctions. Yet proof of research misconduct receives limited attention in the literature. Ethical charters and codes don’t say much about it either. The European Code of Conduct for Research Integrity (All European Academies [ALLEA] 2023: 3.2) states that “persons accused of research misconduct are given full details of the allegation(s) and are allowed a fair process for responding to allegations and presenting evidence”. Likewise, the code of conduct enacted by the German research foundation merely asks research institutions to define appropriate procedures, including “the consideration of evidence” (German Research Foundation [DFG] 2022, guideline 19, explanations). The French Charter of Ethics for Research Professionals (CIRAD, CNRS, INRA, INRIA, INSERM, Institut Curie, IRD, France Université 2015) remains silent on this matter.

To a large extent, proof of research misconduct takes place in a non-judicial context and is decided by people who are not judges, namely the Research Integrity Officers (RIOs) set up in universities or research organisations, or competent at national level. When they receive a report of suspected research misconduct such as plagiarism, fabrication of data, infringement of authorship, RIOs must first check that the alleged facts may indeed constitute a breach of research integrity. If this is the case, they initiate investigations to examine the facts and finally reach a conclusion on whether or not they are accurate and whether or not they qualify as research misconduct in the light of any national legislation developed in this area and the standards generally accepted in the scientific community. The investigations conducted by RIOs being non-judicial, national evidence law does not apply; the rules of evidence before them are prescribed by procedures internal to the institution and subject to local variations.

On the basis of the facts found by the RIO, it is then up to the head of the organisation employing the alleged perpetrator to decide whether or not to initiate disciplinary proceedings before a competent body. The university may also, alternatively or cumulatively, refer the matter directly to the civil courts (to seek compensation) or the criminal courts (to obtain a criminal sanction). The assessment of the facts by the RIOs will determine the starting point for calculating the limitation periods to which these actions are subject under national law. In all these instances, whether civil, criminal, or disciplinary, or in any appeals against the decisions taken in the first instance, the evidence must be re-examined by the court, which is not bound by the conclusions of the RIO. In court, however, proof must be provided in accordance with national law of evidence, failing which the disciplinary or judicial decision may be quashed. In addition, States that are bound by the European Convention on Human Rights must comply with the requirements arising from the Convention. While the latter “does not lay down any rules on the admissibility of evidence as such, which is therefore primarily a matter for regulation under national law”,¹ it does guarantee the right to a fair trial and requires cases to be heard within a reasonable time (Brems and Gerards 2014; Jackson and Summers 2018; Hirvelä and Heikkilä 2021; Jämsä 2023).

Seminars held by the US National Academy of Sciences in the early 1990s highlighted the existence of differences of opinion between government agencies and American universities on the presumption of innocence, the applicable standards of proof and the requirements of due process (National Academy of Sciences, National Academy of Engineering, Institute of Medicine [NASEM] 1992: 112sq.). The fundamental question underlying these differing views is whether RIO investigations are a continuation of scientific discussion between peers (and therefore governed by scientific proof) or whether they should already be considered as constituting a trial (and therefore governed by the law of evidence and the requirements of a fair trial) – or whether they fall somewhere in between. Paul Friedman noted in 1993 that the investigation of research misconduct is “oriented towards information-gathering, analysis, and conclusions. Instead of a jury, there is a panel of faculty members, who act as investigators and judges. Formal evidentiary rules are not followed, but scientific standards of proof are observed. It is much less formal than a trial and offers fewer elements of due process to the accused” (Friedman 1993: s101). This was also lamented by Baltimore, who protested to the director of the then US Office of Scientific Integrity: “So one’s rights are severely curtailed because we’re not in a trial situation?” (Kevles 1998: 230).

However, although the law of evidence applies only to disciplinary and judicial proceedings and not to RIOs’ investigations, the prospect that the facts established by the RIO may later be used in disciplinary proceedings or a trial before a court prevents RIOs from disregarding the national law of evidence altogether. Proof of research misconduct established by the RIO may weigh heavily on the possibility of proving disciplinary misconduct before the disciplinary boards or a civil fault or criminal offence before the courts. It is therefore essential that the evidence marshalled before the RIO is sufficiently robust to lay the foundation for disciplinary or judicial proceedings. If the facts were established by the RIO in complete ignorance of national law on evidence, this would considerably weaken the proof of misconduct in legal proceedings aimed at sanctioning this misconduct. Notably, the guides for RIOs are much more precise in this respect than the charters and codes on research integrity. For example, the recommendations drawn up by the European Network of Research Integrity Offices state that “detailed and confidential records must be maintained on all aspects of the procedure. Any evidence collected related to the allegation(s) should be gathered in accordance with pertinent local laws/guidelines. Evidence should be protected and secured throughout the process to prevent tampering. Consideration should be given to providing secure access to evidence secured (or copies made) so as to enable continuation of work if appropriate” (The European Network of Research Integrity Offices [ENRIO] 2019). The US legislation also demands that RIOs and universities ensure that evidence is preserved.²

Proving research misconduct therefore requires a continuum of evidence from RIOs to judges. Although they are not under any obligation to do so, RIOs need to anticipate the evidential requirements before disciplinary boards and courts of justice. This is a crucial issue. On the one hand, the international scientific community has strived to develop a shared understanding of research integrity and misconduct. But on the other hand, evidence law in disciplinary, civil or criminal proceedings retains specific national distinctiveness (Lebre de Freitas 2004; Rijavec, Keresteš and Ivanc 2016). There is therefore a risk that the specific features of evidence law in each country will lead to differences in the way scientific misconduct is assessed within universities and research organisations.

This article seeks to clarify this matter by providing a legal analysis of the issues involved in the proof of research misconduct allegedly committed by researchers, university faculty, and students insofar as they are engaged in research activities. It draws on three types of material from French law. On the one hand, it builds on the rules that have been drafted in recent years in France to guide the investigations of the RIOs appointed in universities and research organisations. These rules were first discussed informally within the Réseau des référents à l’intégrité scientifique [Network of research integrity officers] (RESINT) on the basis of hard cases faced by members of the network. They were formalised with the publication by RESINT of a Guide (2018) and later a Procedural Manual (2023). The implementation of these procedures remains discretionary for RIOs, but it might soon become compulsory since the French Office for Scientific Integrity (OFIS) has recently been granted the authority to make binding recommendations to universities concerning the procedures to be followed by RIOs.³ The second material considered in this article is the French law of evidence (Vergès, Vial and Leclerc 2022). This legislation applies to French courts in charge of judging the disciplinary, civil and criminal dimensions of research misconduct. It strikingly also influences RIOs investigations. French evidence law is based on civil law, with an adversarial civil procedure and an inquisitorial criminal procedure—although the differences between these procedures are becoming less pronounced over time (Vergès, Vial and Leclerc 2022). Furthermore, French evidence law is strongly influenced by the European Convention on Human Rights and the interpretation given to it by the European Court of Human Rights. Finally, the third research material consists of a review of disciplinary judgements handed down on appeal by the French Conseil national de l’enseignement supérieur et de la recherche [National Council for Higher Education and Research] (CNESER) between 1991 and 2023 in cases involving academics and students (Leclerc and Klausser 2025). These disciplinary decisions offer vivid insight into the proof of research misconduct in action.

The evidential issues raised by the proof of research misconduct are similar in many countries, although the legal responses given to them may differ. For example, unlike the United States, French evidence law contains no specific provisions relating to standards of proof in civil or criminal matters (Roda 2021). Similarly, under US law, the procedure of discovery makes it possible to access a large number of documents before any trial, whereas many European countries limit the scope of documents that can be requested from the opposing party (Collins 1986; Boucobza 2017). Moreover, each country relies on its own jurisdictions and procedures according to its particular legal culture. However, these and other differences in the legal handling of evidence do not obviate the fact that the different jurisdictions face similar evidentiary issues in proving research misconduct. They relate to the burden of proof, the object of proof, the admissibility of proof and the exclusion of proof. All these aspects are examined in this article. The article concludes by highlighting the differences between evidence before RIOs and before courts and concludes that these roles should not be confused, even if this should result in differences in the way in which proof is made before each of them.

1. Burden of proof

The burden of proof is a procedural arrangement for determining who must convince the judge that the facts alleged by a person are true. Scientific proof demands that the scientist who publishes scientific results should prove their soundness. Accordingly, several research integrity guidelines require scientists to provide proof of their claimed scientific results (ENAI et al. 2019: 18; All European Academies [ALLEA] 2023: 2.7). But the burden of proof in the context of an allegation of a research misconduct differs from that applicable in the context of ordinary scientific activity: proving scientific results cannot be equated with proving allegations of fraud (LaFollette 1992: 167). National laws contain rules designed to allocate the burden of proof between the parties to a trial. These generally differ between civil and criminal proceedings. In civil proceedings, it is generally the party alleging a fact that must prove it (actori incumbit probatio). In criminal proceedings, as in disciplinary proceedings, it is generally the prosecuting party who must prove the guilt of the accused person, who is presumed innocent. While these rules do not, in principle, apply to investigations carried out by an RIO, the distribution of the burden of proof is strikingly similar. The allocation of the burden of proof reflects policy objectives regarding the administration of justice. Specialists in the law of evidence point out that the burden of proof actually covers several burdens: the burden of gathering evidence and the burden of convincing the judge (Taruffo 1992; Damaška 1997; Nance 2006: 186; Hoffschir 2016).

1.1 The burden of production

Primarily, the burden of proof designates the party who must provide evidence to enable the RIO or the judge to decide whether or not the alleged misconduct is proven. This obligation is known as the subjective burden of proof or burden of production. This burden is allocated differently before the RIOs and before the disciplinary, civil and criminal courts (see Table 1):

- Proof that a referral to an RIO is well-founded: the allegation to an RIO that research misconduct has been committed triggers an inquiry phase designed to enable the RIO to assess whether the allegation has substance and if an in-depth investigation is warranted. For this to happen, several codes demand the reporting person to provide the RIO with sufficient evidence (e.g. KNAW, NFU, NOW, TO2-federatie, Vereniging Hogescholen, VSNU 2018: 3; Finnish National Board on Research Integrity [TENK] 2023: 5.2.2; The European Network of Research Integrity Offices [ENRIO] 2019: 18). At this point, the people accused do not have to provide any evidence; they are not even informed that the matter has been referred to the RIO. It is up to the reporting person to provide evidence, either on their own initiative or at the request of the RIO. The reporting person does not have to prove misconduct, but must provide sufficient evidence to suggest that the respondent has behaved improperly and that this behaviour could be counted as a research misconduct. If the evidence provided is insufficient, the RIO has to declare the case closed and cannot open an investigation.

- Proof that the allegation sent to the RIO is well-founded: when the RIO is convinced that the referral is well-founded, an investigation phase opens designed to establish whether or not the alleged facts are proven and to determine whether or not they should be considered a violation of research integrity. At this stage of the procedure, French law does not place the burden of proof on one of the parties in particular, but allows the RIO to take evidence from both. Similarly, in the United States, the law provides on the one hand that “the institution or HHS has the burden of proof for making a finding of research misconduct”;⁴ but on the other hand, “the respondent has the burden of going forward with and the burden of proving, by a preponderance of the evidence, any and all affirmative defenses raised” (§ 93.106(2)) and “any mitigating factors” (§ 93.106(3)). Hence, the burden of providing evidence does not lie solely with the reporting person. The respondent is informed of the investigation and invited by the RIO to respond to the allegations and to provide all evidence to support the allegations. Both the plaintiff and the respondent must provide evidence to strengthen their position. In addition, in France, RIOs do not passively receive evidence presented by them; they must also actively “request communication of documents likely to establish the reality of the matter”.⁵ Similarly, in Switzerland, “the investigating authority shall carry out initial clarifications and adopt the necessary measures to obtain evidence on the basis of the applicable procedural law”.⁶ To this end, the RIO may request expert opinions from third-party scientists, e.g. on plagiarism or data manipulation claims. The RIO can also hear people who might shed light on the alleged misconduct. The RIO forms an opinion based on all the evidence provided by the reporting person and the accused person, as well on the evidence they have gathered on their own initiative.

- Proof that the legal claim is well-founded: Should the head of the university or research organisation decides to impose disciplinary sanctions or to initiate proceedings before a civil or criminal court, the person being prosecuted benefits from the presumption of innocence. It is therefore incumbent on the prosecuting party to prove the alleged disciplinary, civil or criminal wrongdoing. The evidence is not limited to the RIO’s report. The persons accused before the court are invited to discuss each item of evidence presented against them and to provide any evidence likely to convince the judge of their allegations. The judge may also order investigative measures such as calling in outside expertise.

1.2 The burden of persuasion

The burden of proof also indicates which party must convince the judge that the alleged facts are true. Here, the burden of proof is not to determine who must offer evidence, but to decide which party loses the case if the evidence is insufficient. This is known as the objective burden of proof or burden of persuasion. The burden of proof therefore has to do not only with the gathering of evidence but also with the weighing of evidence (Taruffo 2008; Nieva Fenoll et al. 2019: 39; Nieva Fenoll 2023).

As far as the procedure before RIOs is concerned, the burden of persuasion lies with the reporting person. If, after considering all the evidence available, the RIO is not convinced that a violation of research integrity has occurred, the RIO must conclude that there has been no misconduct. Conversely, if the evidence collected is found sufficiently convincing, the RIO must decide that the alleged facts are proven. If the RIO is unable to reach a conclusion, the doubt must benefit the defendant and the RIO has to decide that the alleged facts are not proven. Similarly, in disciplinary, civil, and criminal proceedings, the burden of proof lies with the claimant, i.e. in most cases with the president of the university or research organisation; more rarely, a case can be started by the victim of a research misconduct, for example a victim of plagiarism taking legal action against the perpetrator of the violation. In this case, the person initiating the proceedings bears the burden of convincing the judge, and will lose the case if he or she fails to do so. For example, in a case where a student had been accused of exam fraud, the CNESER ruled that “the fraud had not been proven and Mrs XXX should be acquitted for lack of sufficient evidence”.⁷ Any remaining doubts benefit the accused party (Table 2).

In a few European countries, these classical rules for apportioning the burden of proof are supplemented by a subsidiary rule which requires the court to take account of the availability and ease with which each party to the dispute can provide evidence. Such a rule is, for example, applicable in civil proceedings in Spain (Ley 1/2000 de Enjuiciamiento Civil, Art. 217). Similarly, in Belgium, if the judge considers that, in the specific circumstances of the case, it would be manifestly unreasonable for one party to bear the burden of proof, the judge may, in a motivated judgment, impose the burden of proof on the other party (Civil Code, Art. 8.4). These adjustments to the burden of proof have the advantage of relieving a party of the burden of providing evidence that their opponent could more easily provide, thereby also fostering more ethical behaviour on the part of the parties to the dispute. They are, however, criticised for the risk they may entail of privileging political concerns over the determination of actual facts (Ferrer Beltrán 2019) and are only implemented under exceptional circumstances.

The reference to the conviction of the fact-finder that the alleged facts are or are not proven raises the question of the level of persuasion they must reach. This threshold is called a standard of proof. A standard of proof is a degree of proof of the facts that the judge must consider to have been reached in order to be allowed to decide that a fact is proven. Some countries define the standards of proof applicable at trial, which also apply to proceedings initiated to sanction research misconduct. In the United States, for example, the law provides that “a finding of research misconduct […] requires that (c) The allegation be proven by a preponderance of the evidence”.⁸ The OECD also recommends that “the procedure should identify the minimum burden of proof the investigation must meet when assessing the act and the intent with which it was committed for reaching a conclusion of research misconduct. Typically policies establish that a preponderance of the evidence constitutes a burden of proof for administrative investigations such as those into research misconduct”.⁹ In France, no standards of proof are imposed on judges and RIOs, leaving them with no guidance for their assessment (Vergès, Vial and Leclerc 2022: 106sq.). However, it is clear from certain CNESER decisions that, in disciplinary cases, the CNESER requires judges to be certain of the existence of research misconduct. For example, the CNESER ruled, with regard to a student, that “since the university had not established indisputable evidence against Mr XXX, it appeared to the appeal judges that there was no certainty as to Mr XXX’s guilt; that doubt should therefore benefit the accused”.¹⁰

2. The object of proof

The object of proof is what has to be proven. Here, we need to distinguish between the types of facts that need to be proven and the concrete occurrences of these facts. A first approach might be to say that what needs to be proven is research misconduct, or more precisely the falsification or fabrication of data, plagiarism, or any other questionable research practice depending on the case at hand. From this perspective, proving questionable research practices (QRP) poses far greater challenges than proving fabrication, falsification and plagiarism (FFP). While there is broad agreement on the definition of FFP and their characterisation as research misconduct, the nature of QRPs is more debated. QRPs represent a vast grey area including a variety of behaviours that are deemed more or less problematic. Many of these behaviours are highlighted in codes and reports (Organisation for Economic Co-Operation and Development [OECD] 2007; KNAW, NFU, NOW, TO2-federatie, Vereniging Hogescholen, VSNU 2018; All European Academies [ALLEA] 2023), but the list is by nature open-ended. The identification and characterisation of QRPs involves a much greater margin of uncertainty (Bouter 2024), and may even vary considerably from one discipline to another (Ravn and Sørensen 2021). Consequently, just as it is more difficult to estimate the frequency of QRPs in the absence of agreement on what they cover (Hesselmann, Wienefoet and Reinhart 2014), it may be more difficult to determine what facts need to be proven to substantiate them before an RIO or a court, in the absence of shared agreement on their characterisation. These are serious problems, but they are matter of qualification, not proof. Evidence comes into play to determine the existence of facts that allow misconduct to be qualified.

Therefore, the object of the proof is one or more empirical facts, which can qualify as research misconduct, i.e. from which research misconduct can be inferred (whatever meaning national law gives to research misconduct, and whether it concerns FFP or QRP). Rather than discussing further the difficulties that exist in defining research misconduct, that article examines how to prove the facts that characterise such misconduct. As the decisions handed down by the CNESER show, cases involving breaches of research integrity are often complex and multi-dimensional, involving facts that certainly fall within the scope of research integrity stricto sensu, and others that shift the focus of the dispute to other types of behaviour, such as harassment, violence or negligence, in a context where the accused may also be able to claim extenuating circumstances. It would therefore be simplistic to regard misconduct in research as always referring to clearly identified facts and rules of evidence as mere “epistemic engines” (Laudan 2006).

The facts to be proven are often considered to be material events, located in time and space. This is the case, for example, when it is necessary to prove that a scientist documented in a laboratory notebook an experiment that had not been carried out, or that a scientist included in an article data produced by another scientist without crediting the author. Even when clearly individualised, material facts are not however necessarily easier to prove. For example, the fact that a reviewer has appropriated ideas from a research project or an article that they had to evaluate may be difficult to prove as author ignores the identity of the anonymous referees. This also makes it difficult to prove that a referee has unjustifiably withheld a scientific publication from a rival team. These evidential barriers are extremely difficult to overcome. They are therefore likely to encourage problematic behaviour, as offenders may rightly believe that proving that they have failed to uphold research integrity is at least very difficult, if not impossible (such proof might hence be referred to as probatio diabolica).

However, as Michele Taruffo (1992: 105) has rightly argued, the facts to be proven are not always material occurrences located in time and space. They might be psychic facts, in the sense that it is a question of proving facts that fall within the psychic or volitional sphere: bad faith, error, negligence, intention. In other words, the facts to be proven are either “determined descriptively” or “determined evaluatively” (Wróblewski 1992: 138). This situation is commonplace in proving misconduct in research, where genuine misconduct distinguishes from “sincere errors” (Kohn 1986: 205; LaFollette 1992: 40; Resnik 1998: 81), mere “negligence”, and “recklessness” (Resnik 2017; Caron et al. 2023)¹¹. The assessment of research misconduct therefore requires the identification of material facts, but also proof that these extend beyond simple negligence or sincere error.

Although, as Lex Bouter points out, intentionality is irrelevant for the damage done to the validity of the findings (Bouter 2023), it does remain an issue on which the debate on evidence may turn, both before RIOs and before disciplinary bodies or state courts. This is clearly recognised by ENRIO, which states that “the question of mental state (mens rea) or level of intent should be addressed. Is there a minimum level of intent, e.g. intentionally or by (gross) negligence? A clear standard, or burden of proof, should specify what is required to demonstrate different levels of intent, along with examples of supporting evidence to prove various levels of intent” (The European Network of Research Integrity Offices [ENRIO] 2019).

In proving plagiarism, for example, the proof of material facts of duplication of data is needed as well as proof of the intention to plagiarise, as opposed to clumsiness. The intention to plagiarise is quite straightforward when plagiarism occurs by translation from one language into another, which can hardly happen inadvertently, or when the passages copied are important (LaFollette 1992: 166). However, other situations leave more room for discussion about the degree to which the defendant was aware that he or she was engaging in plagiarism. In fact, according to the CNESER, in France,

there is a difference in nature between plagiarism, which is a disciplinary offence consisting of simply copying entire pages, the assembly of which constitutes the fraudulent body—because it was not produced by the doctoral student himself—of the thesis, and simple errors of methodology from which one cannot deduce a desire to defraud which constitutes a disciplinary offence.¹²

Thus, the CNESER upheld the acquittal of a doctoral student accused of plagiarism, noting that

Ms XXX acknowledges these “methodological errors” in the drafting of her thesis document; that her thesis director gave his assent to her defence and that he, like the rapporteurs, should have detected these defects in the thesis document; Considering therefore that there was no intention on the part of Ms XXX to plagiarise her thesis but rather clumsiness; that her thesis director, due to his other duties, was unable to fully supervise the final drafting of Ms XXX’s thesis document; that the first instance decision against Ms XXX [she had been acquitted at first instance by the university’s disciplinary committee] can only be confirmed.¹³

Conversely, in a case involving an assistant professor who was accused of plagiarism and fraud in the procedure put in place to fill a professorship, the disciplinary court of appeal ruled that the evidence in the case file showed that

the acts of plagiarism of which the respondent is accused are not the result of negligence or a material error which, according to the respondent, would exonerate him from responsibility for their occurrence; that the respondent is guilty of the acts of which he is accused and that he must be punished.¹⁴

In these situations, the object of proof hence becomes more complex. It is necessary not only to prove material facts, but also to convince the judge that misconduct should be inferred from these facts. The debate on the evidence therefore revolves around both the material facts and the inferences that the RIO or the judge is asked to make from them. Thus, a defendant may be prepared to admit the facts but may challenge the inference that the requesting person proposes the RIO or the judge make. For instance, in a disciplinary case in which a student was prosecuted for fraud, the CNESER ruled that

Mr XXX considers that the document found on the examination table after the test does not prove that he used it and that the sanction imposed on him at first instance is disproportionate; that the explanations provided by the deferent have convinced the appeal judges.¹⁵

In that case, the defendant was not contesting the material facts (a document was indeed found on the examination table) but the inference that the judge is being asked to draw from it (it is possible to infer from the presence of this document that the person committed fraud). Similarly, a person accused of plagiarism may admit the existence of passages identical to other published texts or the falsity of certain data, but dispute that it is possible to infer plagiarism or fabrication of data, arguing that it is a matter of simple negligence or an “honest mistake”. The object of the evidence then shifts to the facts that support the proposed inference. What facts can be used to prove that the defendant was not aware that he or she was borrowing data or a text from another person? What evidence is there that the defendant made a mistake without any intention to defraud? To answer these questions, RIOs and judges will have to consider other evidence, such as emails, testimonies from colleagues and recordings, that may shed light on the defendant’s intent.

3. Admissibility of evidence

The admissibility of evidence refers to the evidence that can be presented and taken into account to determine whether or not the alleged facts are proven. Unless otherwise provided by law, all evidence is admissible in court if it is relevant (Thayer 1898: 265), i.e. if it makes the facts to be proven more or less probable. Although this rule has its limits (see sect. 4), proof of research misconduct is subject to the principle of freedom of proof, which allows all relevant evidence to be presented. In accordance to this principle, French law does not limit the methods of proof admissible to prove research misconduct¹⁶ and judges are left free (some would say unguided) to assess the evidence, without the law directing their reasoning (Vergès, Vial and Leclerc 2022: 282). The use of a large number of methods of proof is therefore possible.

The RIOs and judges examine the evidence provided by the reporting person and the respondent, and may take a confession from the respondent. Disciplinary decisions against faculty members demonstrate the importance of testimonies as evidence. These are provided, for example, by colleagues, who relate orally what they have personally seen or heard. Testimonies may also be submitted in writing to the RIO or the judge. The admissibility of a piece of evidence nonetheless depends on internal conditions of validity. For example, the CNESER excludes testimony “that has not been signed or is not accompanied by proof of identity”.¹⁷ Similarly, it ensures that the report of any examination fraud involving students is accurate: the CNESER hence nullified a report “because it is not individualised and does not indicate the time at which the test began and ended, or the elements characterising the fraud”.¹⁸

The personal observations of the RIO or judge also play a major role. The latter examine material evidence and directly observe the existence of fraud, for example unsourced borrowings constituting plagiarism or fraudulent duplication of images in several independent scientific publications. To that end, the RIO or the judge may request access to the laboratory notebook in which the experiments are reported, interviews, photographs, recordings, laboratory computers, or samples (AAAS-ABA National Conference of Lawyers and Scientists 1988: 19; LaFollette 1992: 170). In that case, research materials that were initially collected, filed and preserved for the purposes of research are turned into evidence in disciplinary or judicial settings. Personal observations are mainly performed by RIOs, who have scientific expertise. Judges, who in turn rarely have such skills, and RIOs outside their specialised fields, may call on third parties to examine evidence or determine facts. An expert opinion may be entrusted internally to a member of the RIO’s team, which may comprise specialists from different disciplines. Both RIOs and judges may also commission expert opinions. For example, the CNESER committed experts in graphology to assess whether an accused person had indeed written an insulting letter against the university¹⁹ or to examine frauds in examination alleged against university students.²⁰ In other disciplinary cases, the CNESER ordered an expert in psychiatry to assess the misconduct allegedly committed by a university student²¹ or a faculty member.²²

Before the disciplinary boards in universities, proof of plagiarism is commonly provided by anti-plagiarism software that compares the similarity between two texts and enables boards to adjust the penalty so that it is proportionate to the extent of the plagiarism found. Thus, after noting that the plagiarism rate was 25% in a professional dissertation and 13% in an internship report written by a student, the CNESER considered that “in view of the ratio of plagiarised elements, the sanction of six months’ suspended exclusion from the institution is proportionate to the alleged offences”.²³ The assessment of evidence drawn from anti-plagiarism software raises the question of the threshold at which a rate of similarity suffices to characterise plagiarism. When it finds that the rate is low, the CNESER tends to interpret it as “clumsiness” rather than plagiarism. Thus, in the above-mentioned case where a doctoral thesis was suspected of plagiarism, the doctoral student was acquitted at first instance on the grounds that the “borrowing rate” detected by anti-plagiarism software was only 10%, so that the university’s disciplinary section acknowledged “methodological errors” without indicating a “deliberate desire on the part of the defendant to defraud”.²⁴ However, the use of anti-plagiarism software can be precarious: on several occasions, the CNESER has acquitted students on the grounds of “serious doubt” when the anti-plagiarism software was used improperly by the university²⁵.

4. Exclusion of evidence

Evidence in court results from a balance between epistemic considerations, which argue for the widest possible admissibility of evidence, and legal policy considerations (protection of individuals, secrets, fairness, etc), which argue for the exclusion of certain evidence. The rules of exclusion of evidence are at the heart of this tension in that they lead to the judge being deprived, in the name of the protection of certain interests, of evidence that may have otherwise helped to discover the truth. The balance between these two objectives is a topic of considerable debate among specialists in the law of evidence (Bentham 1827; Taruffo 1992; Damaška 1997; Ferrer Beltrán 2002; Stein 2005). The balance between these two objectives does not lie in the same place, depending on whether the proof is made before the RIO or in court.

Evidence that is a priori admissible, insofar as it is relevant and meets the required standards, may however be excluded from the debate and can be taken into account neither by the RIO nor by the judges. Consequently, the evidentiary discussion before RIOs and judges may concern both the weighing of the evidence and its admissibility. Defendants may demand the exclusion of evidence on several grounds: violation of a protected secret, violation of the fairness of the evidence, violation of the adversarial nature of the proceedings.

4.1 Evidence infringing a protected secret

The evidence presented to prove research misconduct might be claimed to violate a secret protected by law. This may be the case, for example, if the reporting person forwards an email from the defendant to the RIO or the judge. In principle, the defendant is entitled to request that this evidence be excluded, as it infringes the confidentiality of correspondence protected by national law and by the European Convention on Human Rights (Art. 8), provided that the correspondence presented as evidence has not circulated so widely that it may have lost its private nature.²⁶ Similarly, where an alleged infringement concerns a patentable innovation, the respondent could challenge the admissibility of the evidence on the grounds of trade secrets. Finally, it is also conceivable, although less likely in practice, that the exclusion of evidence submitted could be requested on the grounds that it would infringe military secrets. In that regard, the Netherlands Code of Conduct (KNAW, NFU, NOW, TO2-federatie, Vereniging Hogescholen, VSNU 2018: 5.4. 10.b) expressly provides that “the procedure may include provisions stating that, in highly exceptional cases, there may be compelling reasons for components of the research, including data, not to be disclosed to an investigation into alleged research misconduct. Such cases must be recorded and the consent of the board of the institution must be obtained prior to using the components and/or data in question in the scientific research. They must also be mentioned in any results that are made public”.

Before the RIOs in France, whose intervention is non-jurisdictional in nature, these secrets are forcefully imposed. Secrets protected by law preclude the admissibility of certain evidence, thereby giving precedence to the protection of the individual over the ascertainment of the truth. The French RIO Network reminds its members that “the conduct of the investigation procedure must respect the secrets protected by law” (RESINT 2023: 94). Any evidence that infringes a protected secret should be disregarded. Consequently, pursuant to the secrecy of correspondence, an email should not be communicated to RIOs without the consent of its author. RIOs are therefore recommended to ask the person in possession of this evidence to write a statement recounting any exchanges they may have had with the respondent (RESINT 2023: 24). This testimony would be admissible if it recounts an event that the witness personally saw or heard.

The situation changes when proceedings are brought before a disciplinary, civil or criminal court. The protection of secrets in French courts has indeed been notably weakened in recent years, following the recognition by the European Court of Human Rights that the parties to a trial have a right to evidence.²⁷ It follows from the ECHR case law that evidence which infringes a protected secret may nevertheless be declared admissible in court if this is necessary for the exercise of the right to evidence and proportionate. Since the early 2010s in France, this right has been applied to many aspects of evidence, including correspondence secrecy, trade secrets, banking secrets, and attorney-client privilege (Bergeaud 2010; Vergès, Vial and Leclerc 2022: 291; Lhomond 2023). Thus, when presented with evidence that infringes a secret protected by law, a judge may nonetheless decide to declare it admissible if this appears essential to prove legitimate rights and if the infringement of the secret is not excessive. This approach gives the judge a wide margin of discretion. It has led France’s highest administrative court to rule, with regard to disciplinary proceedings, that “when the disciplinary court is presented with documents which it is claimed have been collected in disregard of a secret protected by law, it is only incumbent upon it, after submitting these documents to the adversarial debate, to take account of their origin and the conditions under which they were produced in order to assess, at the end of the adversarial discussion before it, their probative value”.²⁸ The impact of the breach of a secret protected by law hence shifts from the admissibility of the evidence to its weighting: judges must take the breach of a secret into account when determining the probative value of the evidence and not necessarily exclude it from the start.

4.2 Evidence infringing the fairness of proof

Where the disputed evidence is a recording made without the knowledge of the person recorded, the latter may argue that it infringes the principle of fairness in evidence and the right to privacy, both of which are enshrined in French law²⁹ and in the European Convention on Human Rights. The principle of fairness in evidence precludes the admissibility in court of evidence obtained without the knowledge of the person concerned or by staging or ploy: for example, PhD students secretly recording their supervisor to prove harassment, a scientist recording a colleague unknowingly. French law provides in principle that fairness compels the judge to exclude evidence obtained by means of a hidden recording device. However, there is no indication that the same applies under existing law to the investigations of RIOs in France. It is noteworthy that the Procedural Manual designed by the RESINT makes no reference to fairness of evidence. The investigation of RIOs is extra-judicial in nature and the rules relating to evidence in court do not apply to its operations. It can therefore be maintened that evidence obtained in an unfair manner is in principle admissible before the RIO, leaving it to the RIO to determine its probative value.

However, the differences between the investigation of RIOs and courts should not be overstated.

On the one hand, the national law of evidence exerts a considerable influence over the work of the RIOs. In the event that an RIO establishes misconduct on the basis of unfair evidence, subsequent disciplinary proceedings may be precluded on the grounds that the same evidence could then be declared inadmissible before the judge in disciplinary proceedings or in civil or criminal proceedings. Consequently, although the rules on the admissibility of unfair evidence set out in national legislation do not apply to RIOs in a legal sense, they do so in practice, given that the RIOs’ actions are focused on the prospect of possible sanctions in the event of research misconduct where the law of evidence applies.

On the other hand, the evaluation of evidence obtained through unfair means in French courts has recently undergone significant transformations following the acknowledgement of the right to evidence within the European Court of Human Rights (ECHR) case law. The European Court of Human Rights has inferred from the right to evidence that the admission of a surreptitious video recording of an employee may be permissible in a court of law in certain circumstances. ³⁰ This has resulted in the Cour de cassation, France’s highest civil court, determining that evidence obtained through unfair means may nevertheless be admissible in court if the conditions for exercising the right to evidence are met. Specifically, the evidence must be essential and its presentation must be proportionate.³¹ Similarly, the highest administrative court (Conseil d’État) has ruled that disciplinary sanctions may be imposed on a university lecturer on the basis of “the transcript of three telephone conversations recorded, without the lecturer’s knowledge, by one of his students”.³² The charges in this case were sexual harassment, but the ruling has implications for all disciplinary proceedings brought against academics, including those related to research misconduct. In this case, the Conseil d’Etat also observed that the three surreptitious recordings whose admissibility was being challenged had been provided to the disciplinary board of the university by students and not recorded by the university that initiated the proceedings.³³ It was therefore concluded that no unfair manoeuvre could be attributed to the applicant, but rather to third parties to the proceedings. Consequently, while the parties and the judge are bound by fairness in evidence, individuals outside the proceedings are not. As disciplinary proceedings following research misconduct are, by definition, initiated by the university, not all evidence provided to the university by scientists is subject to the requirement of fairness.

4.3 Evidence infringing the adversarial nature of the proceedings

While evidence obtained through unfair means may be admissible in certain circumstances before courts and RIOs, the overall fairness of the procedure must be ensured. The European Convention on Human Rights stipulates that the procedure as a whole must be fair (art. 6), which presupposes that the parties to a trial are informed of the charges against them and are communicated the legal arguments and the evidence of their opponent, so that they can respond and exercise their rights of defence.

In France, this rule holds firmly before RIOs, disciplinary boards, and courts alike. To illustrate, the CNESER overturned on appeal a disciplinary sanction taken by a university against a lecturer on the grounds that “during the first instance proceedings, only incriminating testimonies given by students were heard and none of Mr XXX’s teaching colleagues; as a result, the principle of adversary proceedings was not respected”.³⁴ Although the investigation conducted by the RIOs is not a judicial proceeding, the law requires the RIO to “carry out the necessary investigations in the presence of both parties and may request the disclosure of documents likely to establish the facts” (French Code of research, article D. 211-3, 2°). The Procedures Manual prepared by the French RIO Network (RESINT 2023: 15) elucidates that

this implies, for the respondent, the right (i) to be informed of the existence of a procedure for investigating a report of misconduct concerning him or her, (ii) to be informed of all the documents making up the file and on which the final report will be based, (iii) to be heard, and (iv) to be able to put forward his or her arguments, in writing or orally if debates are organised.

This recommendation is in line with that made by The European Network of Research Integrity Offices [ENRIO] (2019: 15), which states that

Investigation of research misconduct allegations should be conducted in a manner that is fair to all parties and in accordance with relevant laws or regulations. Principles of due process or procedural fairness should be described in policies. Persons accused of research misconduct must be given full details of the allegation in writing and must be afforded a fair process with regards to responding to allegations, asking questions, presenting evidence, calling witnesses (if applicable), and providing responses to information or evidence presented. Witnesses (if applicable) should be accompanied by or seek advice and assistance from anyone of their choosing (e.g. a peer, a lawyer etc).

However, the approach adopted in France is notably flexible, as can be seen from cases in which evidence provided to RIOs and judges were anonymous testimonies. The use of anonymous testimonies offers a degree of protection for witnesses in professional contexts where confidentiality is often limited and where the individuals providing the testimony may be hesitant to speak out due to concerns about potential retaliation. Nevertheless, the use of anonymous testimonies is incompatible with the fundamental tenet that the investigative process, whether conducted before the RIO or in a court of law, must adhere to an adversarial framework. Under French evidence law, the admissibility of anonymous testimonies in a court of law is contingent upon two conditions. Primarily, the identity of the witnesses must be known to the individual presenting the evidence and must be made available to the judge upon request. Secondly, the judge is not permitted to base a decision solely on anonymous testimonies; rather these must be corroborated by other non-anonymous evidence, which must also have been submitted to the opposing party for discussion. This approach, which was initially adopted for evidence in civil and criminal proceedings, has been extended to disciplinary proceedings.³⁵ Anonymous testimonies are now admissible in disciplinary proceedings if sufficiently precise and detailed to allow the accused person to discuss their content effectively.³⁶ Moreover, there has recently been a significant development in administrative case law on this matter, concerning disciplinary proceedings. The Conseil d’Etat has recently acknowledged the legitimacy of a disciplinary authority imposing a penalty upon a civil servant based solely on anonymous testimony, provided that the witnesses express concern that disclosing their identity may result in adverse consequences. In the event that the defendant contests the anonymised testimonies at a subsequent trial, the disciplinary authority is obliged to “produce all evidence to show that the witness’s capacity corresponds to that alleged and all evidence to corroborate the facts recounted in the testimony. The judge’s conviction is determined in the light of these contradictory exchanges, which may be supplemented, in case of doubt, by ordering any useful investigative measure”.³⁷

5. Conclusion

This article emphasises the significance of the rules of evidence as a crucial element in the enforcement of research integrity policies. It is inevitable that interactions will occur between the research culture and the judicial culture prevailing in a given context, although this is often overlooked. Restrictive rules of evidence, which impede the discovery of facts and permit unfair conduct by parties to proceedings, can reinforce an unhealthy research culture and foster a sense of impunity and, ultimately, moral disengagement (Redman and Caplan 2017) among scientists. Conversely, rules of evidence that favour the determination of facts and contain few exclusionary rules may prove more effective in establishing breaches of research integrity. In this regard, the European Court of Human Rights’ recognition of a right to evidence can be regarded as a positive development, assuming that the admissibility of more evidence, however illicit or unfair, is generally conducive to the ascertainment of accurate facts in a court of law. Moreover, it is imperative that the procedure for substantiating misconduct be adversarial. The evidence gathered must be made known to the adversary, and both parties must be afforded the opportunity to respond to it and provide evidence in support of their own assertions. Thus, it is not only the fairness of the evidence that is necessary, but also the fairness of the procedure as a whole.

Nevertheless, the tension between rules of evidence that favour the discovery of facts and rules of evidence that excludes certain pieces of evidence on the grounds that they are unfair or infringe certain rights does not manifest in the same way before the RIO and the courts. The recognition of the right to evidence by the European Court of Human Rights and national courts has accentuated the discrepancy between the nature of evidence presented before the RIOs and that presented before the courts. As the right to evidence derives from the right to access a judge that the ECHR infers from article 6 of the European Convention, this right can only be applied before a court and not before the RIO. This results in a significantly broader scope for the evaluation of evidence in disciplinary and judicial proceedings than before RIOs. In the context of disciplinary, civil, and criminal courts, recent developments in French and European law have prompted a greater reluctance concerning the exclusion of evidence. This illustrates a shift in favour of uncovering the truth, even if, to do so, it becomes necessary to declare as admissible evidence that infringes the rights of the person against whom it is invoked. This is not the case with regard to the RIOs, which are unable to declare evidence admissible if it infringes a protected secret. Consequently, the rules of evidence applicable before the courts may permit the admission of a broader range of evidence than before RIOs. These differences illustrate a paradox. Indeed, the balance between epistemic considerations (the orientation of the rules of evidence towards the discovery of the truth) and legal policy considerations (the orientation of the rules of evidence towards the protection of individual or collective rights) is more favourable to the discovery of the truth before judges than before RIOs. This finding may appear surprising given that RIOs are typically scientists evaluating scientific practices. This could suggest that epistemic considerations should take precedence over legal policy considerations and that evidence should be declared admissible more broadly.

Are these differences warranted or should they be reduced or even eliminated? Should RIOs be brought closer to judges and should the rules of evidence applicable before them be aligned with those applicable in the courts? With regard to the presentation of evidence, there is a certain appeal in aligning the status of RIOs with that of judges. This would permit the submission of a greater quantity of evidence and the transformation of their investigations into a quasi-judicial “court of science”. However, beyond the question of evidence, there are significant disadvantages to equating RIOs with judges. RIOs are responsible for evaluating conduct in accordance with the established norms and values of research integrity. In contrast, judges consider a broader range of factors, including the interests at stake, the individual involved, the material conditions in which the individual carries out their activities, and the potential influence of third parties or the institutions within which the scientific activities are conducted. Consequently, if we wish to permit RIOs to conduct investigations on the basis of scientific criteria alone, it is essential to accept the concomitant limitations, including with regard to the proof of research misconduct.

Funding

This work was supported by the French Agence Nationale de la Recherche [grant number: ANR-20-CE27-0016].

Notes

References