dissenting.
The majority holds that the petitioner was not denied due process because (1) the record contains sufficient indicia of the reliability of the confidential informants; (2) the petitioner waived his right to receive any exculpatory material by not appearing at the disciplinary proceedings; and (3) the magistrate properly determined that the petitioner’s counsel should not be permitted to review the in camera material. Although I have reservations about the majority’s analysis of the last two issues,1 I dissent only as to the first.2 In my view, the administrative record and the in camera submission fail to demonstrate that the Institutional Disciplinary Committee (“IDC”) made a bona fide reliability determination or to support the reliability determination apparently adopted by the IDC. I would therefore reverse the grant of summary judgment.
I
The majority acknowledges that this circuit has held that in order to protect the *1300inmate’s right to a fair hearing, when prison disciplinary action is based primarily upon information provided by confidential informants, the record must contain some indication of the reliability of the informants. See Dawson v. Smith, 719 F.2d 896, 899 (7th Cir.1983), cert. denied, 466 U.S. 929, 104 S.Ct. 1714, 80 L.Ed.2d 186 (1984); Jackson v. Carlson, 707 F.2d 948, 948 (7th Cir.), cert. denied, sub nom. Yeager v. United States, 464 U.S. 861, 104 S.Ct. 189, 78 L.Ed.2d 167 (1983); McCollum v. Miller, 695 F.2d 1044, 1048-49 (7th Cir.1983). As the Eleventh Circuit recently observed, the purpose of this requirement is to ensure that prison inmates receive a genuine fact-finding hearing in which the finding of guilt is not based solely on uncorroborated, unreliable hearsay. Kyle v. Hanberry, 677 F.2d 1386, 1390 (11th Cir.1982). “Consequently, to make a decision based on the factual evidence presented, part of the disciplinary committee’s task must be to make a bona fide evaluation of the credibility and reliability of that evidence.” Id. The task of the reviewing court is to examine the record to determine if the committee made such a bona fide evaluation. Dawson, 719 F.2d at 899. It does not review the record de novo, nor can it supply post hoc rationalizations when the IDC either fails to set forth sufficient evidence to demonstrate that it made such an evaluation or it gives grossly inadequate reasons.
II
The majority holds that, when the IDC is unwilling because of institutional safety concerns to document on the record the confidential information and the factual bases of its reliability determinations, the IDC may demonstrate that it did not act in an arbitrary and capricious manner by relying on confidential information that bore no indicia of reliability in any one of four ways:
(1) demonstrate that the investigating officer swore to the truth of the report containing the confidential information and that he appeared to testify before the disciplinary committee (citing McCol-lum, 695 F.2d at 1049);
(2) demonstrate that there was sufficient corroborating testimony (citing Jackson, 707 F.2d at 948);
(3) identify a statement on the record by the chairman of the disciplinary committee that, “he had first-hand knowledge of the sources of information and considered them reliable on the basis of their past record of reliability” (citing Jackson, 707 F.2d at 948); or
(4) submit for in camera review material documenting the investigator’s assessment of the credibility of the confidential informant (citing Dawson, 719 F.2d at 899).
See ante at 1293.
Apparently conceding that the IDC is unable to make the required showing under methods (l)-(3),3 the majority turns to the *1301fourth method, in camera review of a confidential report prepared by the prison investigator4 who compiled the investigative report upon which the IDC relied. In so doing, the majority rejects the arguments of the petitioner that the IDC must make the reliability determination itself, i.e., it cannot adopt the finding of a prison investigator, and that the IDC must include the factual bases for that determination in the administrative record.5 Although I believe that the petitioner’s arguments carry substantial weight,6 I do not address them because I would find that even under the approach adopted by the majority7 the IDC has failed to demonstrate that it undertook a bona fide evaluation of the informants’ reliability.
A
In Dawson this court held that the IDC demonstrates that it undertook a bona fide evaluation of the confidential informant’s reliability (or credibility) when the in camera material “contains more than sufficient additional information to bolster the reliability of the [confidential] information,” id. at 899, and when it shows that the IDC *1302“adopt[ed] the credibility determination made by the prison investigator,” id. Because the court in that case did not discuss the contents of the in camera submission, it is unclear what documentation of reliability the court believed was sufficient to “provide ... [the petitioner] with due process.” Id.
It is clear, however, after Ponte, — U.S. -, 105 S.Ct. 2192 (1985), that the in camera submission must document with specificity the prison investigator’s basis for concluding that the confidential information was indeed reliable in the particular case under investigation, and it must show that the prison investigator’s conclusion was “reasonable,” see id. at-, 105 S.Ct. at 2196 (judicial review ensures that prison officials’ reasons for refusing to call one of inmate’s proffered witnesses are “logically related to preventing undue hazards”); id. at-, 105 S.Ct. at 2201 (Marshall, J., dissenting) {in camera submission must contain the specific (contemporaneous) reasons why prison officials denied the prisoner the rights to call his witnesses so that reviewing court can decide if the prison officials’ decision was based on permissible factors); see also Kyle, 677 F.2d at 1290; Helms v. Hewitt, 655 F.2d 487, 502 (3d Cir.1981), rev’d on other grounds, 459 U.S. 460, 103 S.Ct. 864, 74 L.Ed.2d 675 (1983); cf Aguilar v. Texas, 378 U.S. 108, 84 S.Ct. 1509, 12 L.Ed.2d 723 (in determining if an informant’s statements may form the basis for a determination of probable cause the record must contain sufficient factual evidence to show that the tribunal could reasonably determine that the informant was believable and that the information provided it was itself accurate), particularly in cases such as the instant one where the in camera submission is the only document that purportedly contains the information necessary for a reviewing court to sustain the prison officials’ reliability determination. A reviewing court simply cannot determine “whether the finding of guilt was based on substantial evidence or whether it was sufficiently arbitrary so as to be a denial of the inmate’s due process”, Chavis v. Rowe, 643 F.2d 1281, 1287 (7th Cir.), cert. denied, 454 U.S. 907, 102 S.Ct. 415, 70 L.Ed.2d 225 (1981), if the in camera submission does nothing more than make conclusory assertions regarding the informants’ reliability. Nor can the court sustain the prison disciplinary action as “non-arbitrary” if the prison investigator’s determinations of credibility are unreasonable.
The requirement of specificity does not jeopardize institutional safety. The prison investigator and the IDC are well aware that the confidential report will never be made public. This requirement will also protect against the very real danger of inmate contrivance and resulting inaccurate determinations of guilt. See McCol-lum, 695 F.2d at 1049. In addition, although the requirement does place a greater burden on prison officials, prison officials have already determined that this burden is not onerous and that it necessarily must be undertaken in order to ensure effective judicial review. See infra note 10.
B
The majority finds that the in camera submission in this case “contains more than sufficient additional information to bolster the reliability of the confidential informa-tion_” Ante at 1296. I disagree. The prison investigator’s in camera report merely identifies the confidential informants and state that all of these informants have previously given correct information (presumably to prison officials). The prison investigator only specifically identified one wholly unrelated incident in which less than all of the informants had given correct information. Thus, there was absolutely nothing in that report that suggested why the informants’ identification of the petitioner should be believed or disbelieved in this particular case. In fact, some information given by the confidential informants was directly contradicted by specific evidence that was later uncovered by the prison investigator and that was eventually revealed to the petitioner.
*1303The in camera submission also does not set forth any facts from which we can conclude that the IDC adopted the credibility determination made by the prison investigator. See ante at 1296. The only evidence in the record that indicates that the IDC accepted the credibility determination made by the prison investigator was the incident report in which the IDC stated that “[e]xtensive investigation was conducted by ... [the prison investigator] ... and said investigation contained information from confidential and reliable sources ...” The incident report revealed however that the IDC’s findings regarding Mendoza’s guilt was based on the investigative report; the investigative report, as the majority notes, is a comprehensive thirty-one page document, one and one-half pages of which is devoted to detailing the information uncovered by the prison investigator regarding the petitioner’s alleged participation in the murder. But nothing in the administrative record shows that the confidential report was part of the investigative report.
In addition, the confidential information contained in the investigative report and the in camera submission differed significantly. In the investigative report, the confidential informant’s detailed the petitioner’s relationship with the victim and his participation in carrying out a “contract” on the victim. In the in camera submission, the confidential informants only identified the petitioner as the killer. Thus, it is not even clear that the confidential informants found to be reliable by the prison investigator are the same sources whose information formed the basis of the investigative report. And there is no suggestion in the investigative report that the confidential sources whose information formed the basis for that report were reliable. Nor was the investigator called as a witness at the hearing nor did he swear to the truth of the report. As this court recognized in McCollum, the “investigative report, however vivid and apparently true, is not, ... self-validating.” 695 F.2d at 1049. There is simply no evidence that the prison investigator’s credibility determination was adopted by the IDC.
The statement in the incident report that “[confidential sources are known by this chairman to be reliable” adds nothing to this analysis. There is nothing in the record that shows the chairman’s factual basis for this determination. The chairman might have his own personal source of knowledge, or he may have been merely making a conclusory assertion based upon his prior dealings with confidential informants within the system. It cannot even be inferred from this statement that the chairman was relying on the prison investigator’s determination.
Thus, there are simply no facts contained in either the incident or in camera reports from which we can determine that the prison investigator made a reasonable reliability determination or that the IDC adopted the credibility determination of the prison investigator. See Gomes v. Travisono, 510 F.2d 537, 540 (1st Cir.1974) (“If the written statement is intended to withstand scrutiny and guard against misunderstanding it cannot indicate reliance on speculation or facts not in the record.”).
Finally, the prison investigator’s affidavit, submitted to the court long after the IDC hearing, does not cure the insufficiency of the incident and in camera reports. Although Ponte v. Real, — U.S. -, -, 105 S.Ct. 2192, 2196, 85 L.Ed.2d 553 (1985), suggests that reasons for a denial of a petitioner’s constitutional rights may be given for the first time during habeas corpus proceedings,8 those reasons, of course, cannot be constitutionally infirm. The affidavit presented here contains only the same conclusory assertions contained in the in camera report. The only difference *1304between the affidavit and the in camera submission is that in the affidavit the prison investigator has sworn to the truth. This post hoc attempt to validate the investigative and in camera reports does nothing to show that any proper determination of credibility had been made at the time of the hearing. Furthermore, although the prison investigator stated in the affidavit that he had given the IDC chairman his handwritten material identifying the informants and incorporating his material documenting the informants’ reliability, nothing in the affidavit indicates that the handwritten material documented the reliability of the informants in any greater detail than did the in camera report.
In this case, then, there has simply been no showing that the prison investigator submitted to the IDC anything other than a conclusory finding of reliability that might have been adopted by the IDC. See Kyle, 677 F.2d at 1890-91. (“The government argues that it is enough for the IDC to know that the informant has a past record of reliability. That knowledge, however, is not on the record here. Both the investigator’s report and the confidential report indicated that their sources were considered reliable, but neither one explained why. Indeed, there is nothing on the record to show that the IDC made any inquiry into reliability or that it was furnished with any information explaining the trustworthiness and credibility of the ‘reliable sources.’ ”).
Ill
The majority thus concludes that the only indication of reliability that prison officials need set forth is an unsworn in camera report that reveals the names of the informants and asserts that the informants have given “reliable information in the past” in a wholly unrelated incident.9 This is simply an insufficient basis, however, from which this court can conclude that the IDC made a bona fide evaluation of the reliability of this evidence.
More importantly, the majority’s result today conflicts with Wolff v. McDonald, 418 U.S. 539, 564, 94 S.Ct. 2963, 2978, 41 L.Ed.2d 935 (1974). In Wolff the Court enumerated four procedural safeguards that it believed were essential to secure, at a minimum, some degree of procedural due process for the inmate. That enumeration, however, did not foreclose the “possibility that other procedures besides those specifically considered there may be required to ensure that a disciplinary committee’s actions comport with due process.” Kyle, 677 F.2d at 1389. But see Dawson, 719 F.2d at 899 (suggesting that only those procedures specifically mandated by the Supreme Court are required in prison disciplinary proceedings). In fact, Wolff specifically contemplated under its “mutual accommodation” principle that the inmate would be entitled to all facets of due process that are feasible within the security limitations of penological institutions. As this court recognized in McCollum and subsequent cases, one important element of due process is a decision that is not based on uncorroborated, unreliable, hearsay evidence. Although this circuit has suggested that reliability determinations at the United States Penitentiary at Marion are infeasible, see Dawson, 719 F.2d at 899, prison officials have decided, at least at the time of the hearing in this case, that detailed statements of facts (set forth in the administrative record) supporting the IDC’s reliability determinations are in fact feasible at that institution.10 See McCollum v. *1305Miller, Nos. 81-C-4157, 81-C-4208, 81-C-4172, 81-C-4237, Memorandum and Order at 4-5 (S.D.Ill. May 8, 1985). Thus, under Wolff, due process would seem to require more detailed reliability determinations than were made in the instant case. Therefore, we simply cannot uphold the prison disciplinary action taken in this case based upon the conclusory assertions of reliability made by the IDC and the prison investigator.
As the Supreme Court stated in Ponte: [T]o hold that the Due Process Clause confers a circumscribed right on the inmate to call witnesses at a disciplinary hearing, and then conclude that no explanation need ever be vouched for the denial of that right, either in the disciplinary proceeding itself or if that proceeding be later challenged in court, would change an admittedly circumscribed right into a privilege conferred in the unreviewable discretion of the disciplinary board. We think our holding in Wolff, supra, meant something more than that.
Id. at-, 105 S.Ct. at 2197.
The Due Process Clause confers on the inmate a right to a fair hearing and a decision by the factfinder, based on competent and reliable evidence. In this circuit, the IDC need now provide no rationale for its decision other than a conclusory assertion in an unsworn in camera report that confidential reliable informants identified the inmate as the culprit. As a result, the prison inmate’s rights to a fair hearing and a non-arbitrary decision have been transformed to “privilege[s] conferred in the unreviewable discretion of the disciplinary board.” McCollum’s reliability requirement is effectively eviscerated and Wolff’s holding that prison inmates have due process rights in prison disciplinary proceedings has been essentially gutted.
I would reverse the magistrate’s grant of summary judgment for the defendant and remand, with directions that the IDC be ordered to conduct a disciplinary hearing that meets constitutional standards of due process.
. In particular, I believe that some parts of the in camera submission should have been released to the petitioner and his counsel since it contained information that was revealed to the petitioner before or at his criminal trial.
It is also questionable whether the Institutional Discipline Committee ("IDC”) has provided a "written statement by the factfinders as to the evidence relied on and the reasons for the disciplinary action." Wolff v. McDonnell, 418 U.S. 539, 564, 94 S.Ct. 2963, 2978, 41 L.Ed.2d 935 (1974), quoting Morrissey v. Brewer, 408 U.S. 471, 489, 92 S.Ct. 2593, 2604, 33 L.Ed.2d 484 (1972). The "incident report” of the IDC reads:
Extensive investigation was conducted by Mr. Christie, SIS [special investigative supervisor] and said investigation contained information from confidential and reliable sources which identified inmate Mendoza as one of two inmates that killed inmate Cowger. This occurred at approximately 5:20 p.m. 3/14/81, inmate Mendoza and another inmate repeatedly stabbed inmate Cowger in F-A-l Cowger’s cell. Confidential sources are known by this Chairman to be reliable. This incident report was delayed because of FBI investigation.
This one paragraph, standing alone, clearly is not the summary of reasons and evidence that Wolff requires. The statement that "confidential and reliable sources identified the petitioner as one of the killers” is a mere conclusion at best.
The incident report, however, also indicates that the IDC’s decision was based on a report prepared by a prison investigative supervisor. That report, which is part of the administrative record, contains information culled from confidential sources by the prison investigator, and it recites in some detail facts from which it could be inferred that the petitioner participated in the murder. It can be argued, therefore, that by incorporating by reference the prison investigator's report, the IDC has satisfied this requirement. But this circuit, at least implicitly, has rejected the view that incorporation by reference is sufficient to satisfy Wolff. See Hayes v. Walker, 555 F.2d 625, 631 (7th Cir.1977) ("Hayes /”) (finding that nearly identical statement of reasons and evidence “did not meet minimum due process requirements as set forth in Wolff"); see also Hayes v. Thompson, 637 F.2d 483, 489 (7th Cir.1980) (reaffirming Hayes I since nothing new was introduced on remand that cured constitutional infirmity of original statement of reasons); Aikens v. Lash, 514 F.2d 55, 60-61 (7th Cir.1975), vacated on other grounds, 425 U.S. 947, 96 S.Ct. 1721, 48 L.Ed.2d 191 (1976), reinstated as modified on other grounds, 547 F.2d 372 (1976) (written statement must contain "findings of fact and conclusions based on substantial evidence”). Thus, the IDC’s “written statement” is defective.
. I agree with the majority that the petitioner was not entitled to receive the names of the confidential informants, either before or after the hearing, and that he waived his right to be notified in advance that confidential information would be relied upon. The nature of the confidential informants assertions would have been revealed to the petitioner if he had requested staff representation.
. The administrative record clearly does not contain the requisite showing of a bona fide evaluation. The incident report contains nothing more than a conclusory assertion of reliability. The investigative report, upon which the IDC relied, does not itself contain the requisite showing. There is no assertion in that report that the informants were reliable. See Dawson v. Smith, 719 F.2d 896, 899 (7th Cir.1983), cert. denied, 929 U.S. 466, 104 S.Ct. 1714, 80 L.Ed.2d 186 (1984) ("The prison investigator’s incident report specifically recites that the source of the confidential information was ‘considered to be reliable.’ ”). Furthermore, the investigating officer did not swear to the truth of that report nor was he called as a witness at the disciplinary hearing.
The corroborating testimony of Kaffenberger is also insufficient to demonstrate that the informants were reliable. That statement was only minimally probative of whether the petitioner may have been involved in the incident, because Kaffenberger trapped three other inmates "out of bounds” and because Kaffenberger initially identified Fiaalii as the other inmate involved in the incident. It is decidedly less corroborative than the statement made by the prison inmate in Jackson v. Carlson, 707 F.2d 943, 948 (7th Cir.), cert. denied sub nom. Yeager v. United States, 464 U.S. 861, 104 S.Ct. 189, 78 L.Ed.2d 167 ( 983), who expressly identified the petitioner in that case.
Finally, there is no statement in the record by the chairman that satisfies the third method of demonstrating that the IDC made a bona fide evaluation of reliability.
. Only one of the documents, the prison investigator's July 18, 1981 memorandum, in the in camera submission is relevant to the analysis of whether the prison investigator made a reasonable reliability determination at the time of the hearing. The other documents contain information that was obtained long after the disciplinary proceeding. It is axiomatic that a reviewing court cannot uphold the decision of an agency on the basis of facts not contained in the record or post hoc rationalizations.
. Petitioner’s argument is not limited to a claim that the IDC failed to place the names of the confidential informants in the administrative record. There is no doubt that the petitioner is not entitled to this information. But that does not mean that the factual basis for the determination of reliability should not be placed somewhere in the administrative record. See Kyle v. Hanberry, 677 F.2d 1386, 1390 n. 2 (11th Cir. 1982).
. The very purpose of requiring a disciplinary hearing is to ensure that the IDC does not make arbitrary assessments of guilt. That purpose is best fulfilled if the IDC, the factfinder, makes the reliability determination itself and is required to include the necessary facts for its determination in the administrative record. This procedure ensures the most thoughtful and detailed decisionmaking and affords the prison inmate a basis for challenging inaccurate determinations of guilt. In addition, judicial review is more meaningful when based on a complete and accurate record compiled before the habeas corpus proceedings have begun.
Ponte v. Real, — U.S.-, 105 S.Ct. 2192, 85 L.Ed.2d 553 (1985), does not support the result reached by the majority. In Ponte the Court held that a prison disciplinary board was not required to place in the administrative record the reasons why it had refused to allow the inmate to call one of his proposed witnesses. In so doing, the Court expressly recognized that in Wolff it had placed specific parameters on the prison inmates rights to call and present witnesses and that prison officials were allowed considerable latitude in deciding when to permit prison inmates to call witnesses. Id. at -, 105 S.Ct. at 2197. See Wolff, 418 U.S. at 566-67, 94 S.Ct. at 2979-80. However, it is clear that in Wolff the Court simply did not address what other procedures were constitutionally necessary as a check on the use by prison officials of confidential, and perhaps unreliable, prison information.
Furthermore, the Court in Ponte suggested that the prison officials’ reasons for refusing to call the witness should, in most cases, be revealed in a public record. See id. at-, 105 S.Ct. at 2196 (“We think the answer to that question is that prison officials may be required to explain, in a limited manner, the reasons why witnesses were not allowed to testify, but that they may do so either by making the explanation a part of the ‘administrative record’ in the disciplinary proceeding, or by presenting testimony in court if the deprivation of a ‘liberty’ interest is challenged because that claimed defect in the hearing.”) (emphasis added). Similarly, absent compelling circumstances which the IDC bears the burden of proving, see Ponte, — U.S. at-, 105 S.Ct. at 2199-2200 (Stevens, J. concurring); Wolff, 418 U.S. at 565, 94 S.Ct. at 2979, the prison officials’ factual basis for accepting the confidential sources of information should appear in a public record.
Further support for the petitioner’s argument can be found in the guidelines governing disciplinary hearings issued by Marion prison officials. See infra note 10. Those guidelines mandate, to the extent feasible, that the safeguards advanced by the petitioner here be followed in all disciplinary proceedings. Because Wolff requires that an inmate be afforded all due process safeguards that are feasible under the circumstances, these safeguards are required.
.My analysis of this case under the majority’s approach should not be considered as an endorsement of the four-part test adopted by the majority.
. It is unclear from Ponte whether the Court intended to permit the prison officials to submit post hoc rationalizations. Justice Marshall assumed in his dissent that the Court was validating post-hoc rationalization presumably because if the reasons were not contained in the administrative record in the first instance a court would never know if the officials were offering post-hoc rationalization.
. The majority aptly notes that my difference with this result "invites a response detailing the specifics of the reports," see ante note 4, which it cannot undertake for fear of revealing certain confidential information. Under the in camera method of reviewing prison disciplinary proceedings adopted by this circuit, majority (and dissenting) opinions are necessarily drafted in vague, inconclusive language that gives the petitioner no hint of the identity of his accusers or the extent of the evidence marshalled against him. Thus, prison disciplinary proceedings are now akin to "secret trials”; to the petitioner’s continued protestations of innocence, first the IDC, then the district court, and finally the court of appeals answer that, based on secret knowledge available only to them, he is guilty.
. Relevant sections of Policy Statement 5270.5 provide:
2. BACKGROUND: The Inmate Discipline Program Statement requires that ... IDC de*1305cisions be based on substantial evidence.... Proper documentation of the information utilized in making these decisions is imperative for the initial disciplinary action as well as for later administrative or judicial review.
4. GUIDELINES: When a disciplinary committee decision is based on confidential informant information, ... IDC shall state, on the hearing record, its finding as to the reliability of each informant relied on and the factual basis of that finding.
The reliability of an informant must be established before the information provided may be used to support a finding by the ... IDC. Reliability may be determined by a record of past reliability or by other factors which reasonably convince the ... IDC chairman of the informant’s reliability. The staff member providing the information to the committee shall include a written statement of the frequency with which the informant has provided information, the period of time during which the informant has provided information, and the degree of accuracy of that information. If reliability is based on factors other than a history of reliability, those other factors supporting a determination of reliability must be clearly specified. Staff have an affirmative obligation to determine whether there is any basis for concluding that the informant is providing false information.
All confidential information presented to the committee shall be in writing and must state facts and the manner in which the informant arrived at knowledge of those facts. If possible, the statement shall be signed by the informant. If the informant does not write a statement, the staff member receiving the information shall provide that information in language as close to the informant’s as possi-ble_ The committee chairman shall include, in the record of the hearing, a statement of the basis for finding that the information provided by the informant is credible.