Charalambos Efstathiadis is a citizen of Greece who emigrated to the United States on or about December 22, 1967, and is a legal permanent resident. On October 19, 2005, Efstathiadis pled guilty to four counts of sexual assault in the fourth degree under Connecticut General Statute § 53a-73a(a)(2). C.G.S. § 53a-73a(a)(2) criminalizes subjecting “another person to sexual contact without such other person’s consent.” “Sexual contact” is further defined as contact “with the intimate parts of a person not married to the' actor for the purpose of sexual gratification of the actor or for the purpose of degrading or humiliating such person.” C.G.S. § 53a-65(3).
On January 7, 2009, the United States commenced removal proceedings against Efstathiadis pursuant to 8 U.S.C. § 1227(a)(2)(A)(ii), which provides that an alien who “is convicted of two or more crimes involving moral turpitude, not arising out of a single scheme of criminal misconduct, ... is deportable.” On April 8, 2009, Michael W. Straus, Immigration Judge, determined that C.G.S. § 53a-73a(a)(2) was not a crime involving moral turpitude (“CIMT”) because “the Connecticut statute does not appear to require that the actor know that his actions were not consented to by the victim.” Oral Decision of the Immigration Judge at 5, Joint App’x *594Vol. 1, at 173 (Dkt. No. 34). The Board of Immigration Appeals (“BIA”) reversed on October 22, 2010, on two alternate grounds. First, the BIA found that C.G.S. § 53a-73a(a)(2) was a CIMT because “the requirement of acting for the purpose of sexual gratification of the actor or an intention to degrade or humiliate the victim presents a realistic probability that the perpetrator had an evil intent.” Decision of the BIA, at 2, Joint App’x Vol. 1, at 128. Second, the BIA applied the Attorney General’s decision in In re Silvar-Trevino, 24 I. & N. Dec. 687 (Nov. 7, 2008), available at 2008 WL 4946455, to go beyond the modified categorical approach and consider the underlying facts of Efstathiadis’ conviction.1 Decision of the BIA, at 2, Joint App’x Vol. 1, at 128. Because the BIA’s decision was non-final, there was a second round of decisions in which, on December 26, 2012, the BIA ultimately applied its 2010 decision as the “law of the case.”
DISCUSSION
I. Jurisdiction & Standard of Review
We have jurisdiction over appeals from “a final order of removal.” 8 U.S.C. § 1252(a)(1), (a)(2)(D). Under the somewhat tortuous jurisdictional provisions of Title 8, however, we lack jurisdiction “to review any final order of removal against an alien who is removable by reason of having committed” certain criminal offenses, including criminal offenses that constitute CIMTs. 8 U.S.C. § 1252(a)(2)(C). Our jurisdiction is reinstated where we are called upon to “review ... constitutional claims or questions of law.” 8 U.S.C. § 1252(a)(2)(D). Put another way, “[w]e retain jurisdiction ... to determine whether this jurisdictional bar applies.” James v. Mukasey, 522 F.3d 250, 253 (2d Cir.2008); see Alsol v. Mukasey, 548 F.3d 207, 210 (2d Cir.2008) (“Thus, we retain jurisdiction to decide the question of law regarding whether this jurisdictional bar applies.... ”). In sum, we have jurisdiction to determine whether C.G.S. § 53a-73a(a)(2) is a CIMT.
“While this Court gives substantial deference to the BIA’s interpretation of the [Immigration and Nationality Act], a statute it is charged with administering, we review de novo its interpretation of state and federal criminal laws.” Santana v. Holder, 714 F.3d 140, 143 (2d Cir.2013); see Weng v. Holder, 562 F.3d 510, 513 (2d Cir.2009) (announcing the same standard of review). Specifically, “[b]eeause the BIA has expertise applying and construing immigration law, we afford Chevron deference to its construction of undefined statutory terms such as ‘moral turpitude.’ However, ... the BIA has no expertise in construing federal and state criminal statutes, and so we review de novo the BIA’s finding that a petitioner’s crime of conviction contains those elements which have been properly found to constitute a CIMT.” Gill v. I.N.S., 420 F.3d 82, 89 (2d Cir.2005) (internal citations omitted).
Because the BIA’s 2012 decision adopted the reasoning of the BIA’s non-final 2010 decision, we effectively review the BIA’s 2010 decision. See Chupina v. Holder, 570 F.3d 99, 105 (2d Cir.2009) (per curiam) (dismissing appeal from non-final order of removal, but noting that exhausted claims would be preserved in a proper appeal when a final order is issued); Weng, 562 F.3d at 513 (“Because the BIA adopted and affirmed the IJ’s decision, we review the two decisions in tandem.”); Chen v. Gonzales, 417 F.3d 268, 271 (2d Cir.2005) (stating the same principle).
*595II. Merits
Whether a prior conviction constitutes a CIMT turns on whether the crime is “inherently base, vile, or depraved.” Mendez v. Mukasey, 547 F.3d 345, 347 (2d Cir.2008). Because “[i]t is in the intent that moral turpitude inheres,” the focus of the analysis is generally “on the mental state reflected” in the statute. Gill, 420 F.3d at 89; see Mendez, 547 F.3d at 347 (“Whether a crime is one involving moral turpitude depends on the offender’s evil intent or corruption of the mind.” (internal quotation marks omitted)).
In making this determination, we apply a categorical approach that “look[s] not to the facts of’ the particular case, “but instead to whether [the offense] categorically fits within” the definition of a CIMT. See Moncrieffe v. Holder, — U.S. -, 133 S.Ct. 1678, 1684, 185 L.Ed.2d 727 (2013). Where the state statute “contain[s] several different crimes, each described separately, ... a court may determine which particular offense the noncitizen was convicted of by examining the charging document^] ... plea agreement^ and] plea colloquy,” among other documents in the judicial record. Id.; see Wala v. Mukasey, 511 F.3d 102, 107-08 (2d Cir.2007) (applying the same standard in a CIMT case). We refer to this as the “modified categorical approach.” See United States v. Beardsley, 691 F.3d 252, 258 (2d Cir.2012); Akinsade v. Holder, 678 F.3d 138, 144 (2d Cir.2012).
Here, there is no dispute that although the statute under which Efstathiadis was convicted — C.G.S. § 53a-73a— contains several subparts, the record clearly establishes his conviction for sub-part (a)(2). C.G.S. § 53a-73a(a)(2), in turn, is informed by the statutory definition of “sexual contact,” which contains two alternate intent elements: acting “for the purpose of sexual gratification of the actor or for the purpose of degrading or humiliating such person.” See C.G.S. § 53a-65(3).
The parties do not address which of these two possible intentions underlie Efs-tathiadis’ conviction. We do not decide the issue because the differences between the categorical and modified categorical approaches have no impact on our determination to certify. Pursuant to either approach, our focus is on the question of whether or not “the minimum conduct criminalized by the statute” would support classification of a crime as a CIMT. Moncrieffe, 133 S.Ct. at 1684; see Gill, 420 F.3d at 89 (“[T]o constitute a CIMT, a criminal category must by definition, and in all instances, contain each of those elements that constitute a CIMT.”). For the reasons we discuss below, we are left uncertain whether the “minimum conduct” necessary to sustain a conviction involving either possible intention supports the classification of C.G.S. § 53a-73a(a)(2) as a CIMT.
C.G.S. § 53a-73a(a)(2) outlines three elements-sexual contact, for a prohibited purpose, and without consent. Connecticut courts have not added a complicating judicial gloss but have applied the statute as written. See, e.g., State v. McGee, 124 Conn.App. 261, 4 A.3d 837, 843 (2010) (reciting the statutory definitions given above); State v. Montoya, 110 Conn.App. 97, 954 A.2d 193, 198 (2008) (same). The Connecticut Criminal Jury Instructions2 *596also identify three elements-sexual contact, “specific intent” either to obtain sexual gratification or to degrade or humiliate the complainant, and “the complainant did not consent to the sexual contact.” Crim. Jury Instructions § 7.1-12. Neither the statute, nor the eases applying the statute, nor the model jury instructions requires any jury finding of mens rea with regard to the lack of consent element.
Generally speaking, mens rea refers to the degree of mental culpability with which a defendant committed the acts underlying a conviction and comes in four basic types (intent, knowledge, recklessness, negligence), with a crime in which mens rea is not required being referred to as a “strict liability” offense. See United States v. Figueroa, 165 F.3d 111, 114 n. 3 (2d Cir.1998) (“The types of mental states required for criminal convictions range from purposefulness (or intention) to knowledge, recklessness or negligence. In some limited circumstances, when the penalties attached to a violation are low and the reputational effects of a conviction are minimal, Congress may also create true ‘strict liability’ crimes, which have no mens rea requirement at all.” (internal citation omitted)).
The level of mens rea attributable to a given act is of vital importance to the culpability we assign to that act. For example, under Connecticut law, to be convicted of either murder or manslaughter a defendant must actually “cause the death” of the victim, but for a murder conviction this act must be undertaken with “the intent to cause the death,” whereas for manslaughter, the “intent to cause serious physical injury” will suffice. Compare C.G.S. § 53a-54a(a) (murder) with C.G.S. § 53a-55(a)(l) (manslaughter). Thus, the mens rea attributable to an act answers the fundamental question of whether the act was intentional, a reasonable mistake, or a mistake that should have been avoided.
Each element of a crime may require a different level of mens rea or none at all. Here, C.G.S. § 53a-73a(a)(2) unambiguously provides for a mens rea component with respect to the sexual contact element by requiring that the contact be with one of two prohibited purposes (intentions). What is unclear, by contrast, is what level of mens rea, if any, applies to the lack of consent element. For example, if the requisite mens rea were knowledge, then the prosecution would be required to prove that a defendant knew that the victim did not consent. If the requisite mens rea were negligence, then the prosecution would need to prove that a reasonable person in the defendant’s position would have known that the victim did not consent, even if the defendant did not actually know. Alternatively, C.G.S. § 53a-73a(a)(2) may not have a mens rea component with regards to lack of consent, i.e., strict liability may attach to that element. In this circumstance, the “minimum conduct” required to sustain a conviction would be sexual contact with the intent to receive sexual gratification, or to humiliate, under circumstances where the victim did not consent, even if the perpetrator reasonably believed that consent were present.
The BIA applied the familiar test that, to qualify as a CIMT, the elements of the crime must demonstrate “an evil intent, or depraved or vicious motive.” Decision of the BIA, at 2, Joint App’x Vol. 1, at 128. *597Our precedents have also emphasized the importance of an evil intent, variously describing the requirement as “a vicious motive or corrupt mind,” or “[a]n evil or malicious intent,” and noting that “corrupt scienter is the touchstone of moral turpitude” and that “it is in the intent that moral turpitude inheres.” Michel v. I.N.S., 206 F.3d 253, 263 (2d Cir.2000) (internal quotation marks omitted) (collecting cases); see Mendez, 547 F.3d at 347 (“Whether a crime is one involving moral turpitude depends on the offender’s evil intent or corruption of the mind.” (internal quotation marks omitted)).
But the intent to receive sexual gratification, standing alone, is not evil. The requisite mens rea, if any, applicable to the lack of consent element is thus relevant to our determination whether C.G.S. § 53a-73a(a)(2) is a CIMT. So, too, might be a circumstance such as the age of the alleged victim.
Nothing in this opinion should be taken to imply that we are holding that negligence as to lack of consent combined with the intent to receive sexual gratification renders C.G.S. § 53a-73a(a)(2) a CIMT. Similarly, nothing should be understood to imply that knowledge of lack of consent is the minimum mens rea required for a CIMT. We expressly decline to reach these questions. We conclude only that in the context of a conviction arising under C.G.S. § 53a-73a(a)(2), where the only factors affecting the moral character of the crime are the intention to receive sexual gratification (or to humiliate) and the mens rea associated with lack of consent, strict liability as to lack of consent would not lend support to categorization as a CIMT, even if we might, on further analysis and other bases, conclude that such categorization were appropriate.3
The Government argues that Connecticut provides for a mistake-of-fact defense that would protect a defendant who reasonably believed he or she had consent. Indeed, sexual assault statutes sometimes do not specify a mens rea as to consent when, in operation, a recognized mistake of fact defense acts as a proxy for mens rea in this context. See Robin Charlow, Bad Acts in Search of a Mens Rea: Anatomy of a Rape, 71 Fordham L.Rev. 263, 276 (2002); see also Peter Westen, The Logic of Consent 147 (2004) (“A lack of ‘wrongful intent’ regarding a victim’s non-consent is a defense of mens rea, and it obtains when a victim’s conduct is such as to cause a defendant reasonably ‘to assume’ that she is voluntarily acquiescing to sexual intercourse in her mind even if she is not.”). We are uncertain, however, as to whether this is the rule in Connecticut. As relevant here, a mistake-of-fact defense applies where a “factual mistake negated the mental state required for the commission of an offense.” C.G.S. § 53a-6(a)(l). So mistake of fact as to consent is a defense if and only if Connecticut requires the prosecution to prove a mental state related to *598lack of consent-the very question we are unable to answer.
The Government also relies on the Connecticut Supreme Court’s decision in State v. Smith, 210 Conn. 132, 554 A.2d 713 (1989). In that case, the Connecticut Supreme Court appeared to speak generally about the nature of consent:
While the word “consent” is commonly regarded as referring to the state of mind of the complainant in a sexual assault case, it cannot be viewed as a wholly subjective concept. Although the actual state of mind of the actor in a criminal case may in many instances be the issue upon which culpability depends, a defendant is not chargeable with knowledge of the internal workings of the minds of others except to the extent that he should reasonably have gained such knowledge from his observations of their conduct.... [W]hether a complainant has consented to intercourse depends upon her manifestations of such consent as reasonably construed. If the conduct of the complainant under all the circumstances should reasonably be viewed as indicating consent to the act of intercourse, a defendant should not be found guilty because of some undisclosed mental reservation on the part of the complainant.
Id. at 717.
Smith, however, considered a prosecution for sexual assault in the first degree, id. at 714, not the fourth, and we are aware of no instances in which Smith has been applied to sexual assault in the fourth degree.4 Moreover, the court’s holding in Smith relied upon features of sexual assault in the first degree that are not applicable to sexual assault in the fourth degree. At the time Smith was decided, sexual assault in the first degree applied “to a person who compels another person to engage in sexual intercourse by the use of force or by the threat of use of force which reasonably causes such person to fear physical injury.” Id. at 716 (internal quotation marks and ellipses omitted). The Connecticut Supreme Court reasoned that “[a] finding that a complainant had consented would implicitly negate a claim that the actor had compelled the complainant by force or threat to engage in sexual intercourse.” Id. at 717; see Crim. Jury Instructions § 7.1 (characterizing Smith by stating that “[cjonsent, often raised as a defense, is more exactly a denial of the element of compulsion”). Sexual assault in the fourth degree requires no analogous force or threat through which a reasonable belief of consent defense could be imputed.
In sum, we are left uncertain about the likelihood that the Connecticut courts would permit a defendant to be convicted under C.G.S. § 53a-73a(a)(2) where the defendant reasonably believed that consent was present. On the one hand, the statute, elements as recited in the cases, and jury instructions suggest strict liability. On the other hand, we are unable to find a case where we are confident strict liability was imposed and the Connecticut Supreme Court has spoken generally to the issue of consent in a manner that suggests strict liability is inappropriate. Even if Connecticut would not apply strict liability to the lack of consent element, we are unable to discern what level of mens rea would be applied.
*599We have the authority to certify an open question of law to the Connecticut Supreme Court. C.G.S. § 51-199b(d) (“The Supreme Court may answer a question of law certified to it by a court of the United States.... ”); Second Cir. Local R. 27.2(a) (“If state law permits, the court may certify a question of state law to that state’s highest court.”). ‘Where a question of statutory interpretation implicates the weighing of policy concerns, principles of comity and federalism strongly support certification.” Parrot v. The Guardian Life Ins. Co. of Am., 338 F.3d 140, 144 (2d Cir.2003) (internal quotation marks omitted). Whether or not Connecticut imposes strict liability for intentional sexual touching without consent implicates important policy concerns, as is clear from the Connecticut Supreme Court’s decision in Smith. Certification, therefore, is appropriate in this case.5
CONCLUSION
Under federal immigration law, Efsta-thiadis’ removal turns on whether the crime he was convicted of — C.G.S. § 53a-73a(a)(2) — is a CIMT. The answer to that question, in turn, implicates the level of mens rea applicable to C.G.S. § 53a-73a(a)(2)’s lack of consent element. Because we are unable to predict what level of mens rea the Connecticut courts would require, and because the issue involves the weighing of important policy considerations, we respectfully certify the following questions to the Connecticut Supreme Court:
(1) Is C.G.S. § 53a-73a(a)(2) a strict liability offense with respect to the lack of consent element?
(2) If C.G.S. § 53a-73a(a)(2) is not a strict liability offense with respect to the lack of consent element, what level of mens rea vis-á-vis that element is required to support a conviction?
Accordingly, it is hereby ORDERED that the Clerk of this Court transmit to the Connecticut Supreme Court a Certificate, as set forth below, together with a complete set of the briefs, appendices, and record filed by the parties in this Court. The parties are further ORDERED to bear equally such costs and fees, if any, as may be required by the Connecticut Supreme Court.
CERTIFICATE
The foregoing is hereby certified to the Connecticut Supreme Court pursuant to Second Circuit Local Rule 27.2(a) and Connecticut General Statute § 51 — 199b(d), as ordered by the United States Court of Appeals for the Second Circuit.
. We do not reach the validity of In re Silva-Trevino because we certify questions to the Connecticut Supreme Court that may, depending on the answers, resolve this case.
. These model instructions are published on the Connecticut Judiciary’s official website and appear to be produced by a committee of judges "as a guide for judges and attorneys in constructing charges and requests to charge and as a general reference to criminal offenses and their elements. The use of these instructions is entirely discretionary and their publication by the Judicial Branch is not a guarantee of their legal sufficiency.” Crimi*596nal Jury Instructions Home, http://www.jud. ct.gov/JI/criminal/default.htm (last visited May 19, 2014).
. By way of example as to factors that, if relevant, would affect our analysis, we note that Efstathiadis was not charged with fourth degree sexual assault involving a minor. See C.G.S. § 53a-73a(a)(l)(A)-(B). We therefore express no opinion as to whether a conviction under that subpart, or a different sexual assault statute involving a minor victim, would constitute a CIMT even in the absence of a mental component as to lack of consent. Cf. Mehboob v. AG of the United States, 549 F.3d 272, 278 (3d Cir.2008) (observing “the consensus that moral turpitude inheres in strict liability sex offenses” involving minors or children (citing cases)); Marinelli v. Ryan, 285 F.2d 474, 475-76 (2d Cir.1961) (“We have no doubt that” C.G.S. § 53-216, the indecent touching of a minor under the age of sixteen, “involves moral turpitude.”).
. We note that the Government does argue that none of the reported cases involving application of C.G.S. § 53a-73a(a)(2) appears to have imposed liability on a defendant who reasonably believed he had consent. But even assuming, arguendo, that the Government is correct, these cases do not discuss whether Smith is applicable in this context.
. While a petition is pending in this Court, the Government's forbearance policy assures that the filing of a motion to stay removal, as has been done here, will suffice to prevent removal. Of course, our decision to certify questions of law to the Connecticut Supreme Court does not alter in any way the fact that Efstathiadis’ petition is pending before this Court. Should the Government disagree, we expect it to inform us promptly so that we might decide the pending motion to stay removal. Cf. In re Immigration Petitions for Review Pending in the U.S. Court of Appeals for the Second Circuit, 702 F.3d 160, 162 (2d Cir.2012).