Human Rights Watch Comments on California Judicial Council Bail Reform Rules

Sent via e-mail and regular mail

Judicial Council of California
Criminal Law Advisory Committee
455 Golden Gate Avenue
San Francisco, CA 94102-3688
Attn.: Eve Hershcopf Eve.Hershcopf@jud.ca.gov 
Kara Partow  Kara.Partow@jud.ca.gov 

RE: HUMAN RIGHTS WATCH’S COMMENTS ON PROPOSED CALIFORNIA JUDICIAL COUNCIL RULES 4.10 AND 4.40

In April 2017, Human Rights Watch published “Not in it for Justice: How California’s Pretrial Detention and Bail System Unfairly Punishes Poor People,”[1] a comprehensive report on the impact of pretrial detention.  We have been deeply involved in advocacy for comprehensive pretrial detention reform in California and throughout the US. 

California’s bail reform law, SB 10, passed despite strong opposition from many organizations who firmly believe in pretrial detention reform, including several who were original co-sponsors of the bill.[2]  Human Rights Watch, in our opposition, raised two primary concerns:[3] 1) it gives too much unchecked discretion to judges without standards to guide them; 2) it requires the use of profile-based risk assessment tools to make cut-off decisions about eligibility for pretrial incarceration and to inform judges’ ultimate decision to release or incarcerate.

Bail reform advocates who remained supportive of SB 10 argued that the rule-making process would help mitigate problems with the law.  Unfortunately, the first two proposed rules from Judicial Council provide little meaningful check on unlimited judicial discretion and no meaningful safeguards against abuse of the risk assessment tools.

Human Rights Watch recommends that the Committee extend the time period to develop these rules and re-write them to address the specific problems with SB 10.

The following are general criticisms of the proposed rules, which are followed by a specific breakdown of each provision:

• The proposed rules provide no definition of the risk categories (low, medium and high) into which risk assessment tools will sort people, including no limitation on what factors the tools will use to make their assessments. 
• The rules allow the tools to assess for improper risks, like risk of missing a court date, instead of risk of fleeing the jurisdiction; or risk of re-arrest for any crime, instead of risk of committing a new dangerous crime.
• The rules allow unproven accusations to be assessed as part of a person’s criminal history, rather than limit it to actual convictions.
• The rules fail to forbid, and may even enable, use of risk assessment tools that lack transparency or produce biased outcomes.
• In areas where real clarity is needed, the rules use vague language and undefined terms rather than give precise guidance on use of the tools.
• The rules do not set procedural requirements to guarantee consistent, careful consideration in decisions by local jurisdictions to create exclusions to release for “medium” risk people and do not provide for input from stakeholders impacted by these decisions.
• The rules do not set meaningful limits on the number of people or categories of people assessed as “medium” risk who can be excluded from release by local courts.
• The rules do not account for the difficulty accused people will have getting favorable information to Pretrial Assessment Services (“PAS”) agents at the pre-arraignment stage.
• The rules do not enhance the due process rights of accused people in deciding their pretrial custody status.

There are some very positive aspects of the rules:

• The rules, at various stages, make clear that courts and PAS must account for the circumstances of the accused individual and the impact pretrial incarceration will have on that person’s family and community.          
• The rules state that conditions of release must not be a form of punishment and must not be made overly difficult to fulfill.

Overview of the Flaws with SB 10 that Judicial Council Rules Should Address

The critique of the proposed rules should be understood in the context of the overall problems of SB 10’s bail “reform” scheme.  The rules should serve to limit the vast judicial discretion to impose preventive detention and to mitigate the harmful impact of profile-based risk assessment tools.

Unlimited Judicial Discretion

SB 10 removes money bail as a mechanism of pretrial incarceration, but it gives judges the power simply to impose direct preventive detention[4] on extremely subjective justifications and without adequate due process protections. 

Section 1320.18 authorizes prosecutors to request preventive detention if one of five circumstances exist, including this catch-all: “if there is a substantial reason to believe that no nonmonetary condition or combination of conditions of pretrial supervision will reasonably assure protection of the public or a victim, or the appearance of the defendant in court as required.” 

The court can then order preventive detention if the court determines by “clear and convincing evidence” that no nonmonetary conditions of release will reasonably assure appearance in court or public safety.[5]  The “clear and convincing” standard is supposed to be more difficult to attain than the “preponderance of evidence” standard, but both standards are subjective.

One of the most significant problems of California existing pretrial detention system is that people in custody plead guilty regardless of actual guilt because it is their quickest way out of jail. Many judges use their discretion to set unattainable money bail, knowing it will pressure guilty pleas.[6]  SB 10 gives judges broad discretion to detain pretrial; these rules do not constrain that discretion.  This broad discretion threatens to simply recreate the same coercive pretrial incarceration system that existed previously.

The detention hearings themselves, described briefly in Section 1320.20(c), lack any rigorous format or due process rules that would enhance confidence in their fairness.  The judge can decide to jail a person pretrial based on offers of proof, arguments of counsel and any “reliable” hearsay.  There are no rules of evidence or right to cross-examination.  These hearings are no more rigorous than the cursory hearings under the existing money bail system.   This lack of process calls for rule-making to impose some requirements, including a right to cross-examination and limits on the use of offers of proof and hearsay, that would regularize the hearings and make them more likely to be fair and less arbitrary.

The judge may also use the risk score generated by the risk assessment tool to justify preventive detention.[7]  The tools are used to decide who is eligible for early release and who remains detained.  Regulation of the use of the tools is, therefore, essential to the fairness of the system.

 

Flawed Risk Assessment Tools

Profile-based risk assessment tools,[8] the type required by SB 10, are inherently flawed.  They make decisions or recommendations not by looking at the overall context of an individual’s situation, including the circumstances of the crime alleged, but by taking certain data points about that individual, creating a profile, and comparing outcomes of a dataset of other people with similar profiles to make a statistical estimate of the likelihood of future events (missing a court date or re-arrest).  This statistical decision-making misses the contextual understanding necessary to insure public safety and to insure the individual receives fair treatment.  For example, a person who missed a court date because he was sick, but came to court the next day, might be scored the same as a person who left the jurisdiction to avoid facing prosecution.

This de-humanizing statistical analysis is made significantly worse because the tools analyze factors that are greatly influenced by racial and class bias.  The tools use arrest history as a critical data-point.  While arrest history, in part, measures the behavior of the individual, it also largely measures the behavior of police—in what neighborhoods they deploy, who they choose to detain, who they choose to search.  There is ample evidence of racial bias, explicit and implicit, in these policing decisions.  Additionally, even if the tools only analyze convictions, bias in court outcomes means that racial minorities and poor people will often receive higher risk scores that do not accurately reflect their relative risk of future dangerousness.  Biased data leads to biased predictions, under cover of the fiction that these tools are making scientific calculations.

The scoring system of these tools can be adjusted to lower, raise or maintain levels of pretrial incarceration.  The primary impetus behind pretrial detention reform was a recognition that too many people were being jailed before ever being convicted of a crime, resulting in lost jobs, damaged family ties, economic harm, personal suffering of the misery of jail, and, perhaps most significant to the credibility of our court system, coerced guilty pleas.[9]  The tools will not necessarily address these identified problems of pretrial detention.

There are other flaws in the risk assessment tools, including their lack of transparency, their inaccuracy and the lack of oversight in their development and implementation.  Effective rule-making might mitigate these flaws.

Judicial Council’s Invitation to Address Key Omitted “Alternatives”

The Judicial Council, in its “Invitation to Comment” on the proposed Rules 4.10 and 4.40, set forth two “alternatives” that they considered but rejected in their proposals.  They asked for input on those alternatives.

Restricted use of risk assessment score

The first alternative was a provision of Rule 4.10 restricting the use of information generated by the risk assessment tool to the pretrial release/detain/conditions of release decision, and not allowing it to be used in subsequent proceedings.  The committee further considered limiting it to use for impeachment purposes only.

Section 1320.9(d) says the PAS report including the risk assessment may not be used for any other purpose than the pretrial release/detain decision, so a separate rule may not be entirely necessary.  However, other aspects of these rules have simply repeated or clarified statutory language, so a rule that removes any ambiguity and forbids subsequent use of the risk score is appropriate.

Given the inherent flaws of risk assessment tools, they should be restricted as much as possible to limit their potential for harm.  If the pretrial risk assessment may be used in subsequent sentencing or probation violation decisions, then its harm is multiplied.  The assessment threatens to become determinative, as judges may disregard mitigation evidence developed by a lawyer during the case if the “scientific” finding is that the person is “high risk.”  Additionally, if risk scores are not sealed immediately after the pretrial custody determination, then they may influence other decisions judges make related to the admission of evidence or holding to answer or approval of plea agreements, as many judges will be hesitant to give a “high risk” person the benefit of the doubt in judgement calls.

Definition of “criminal history”

The second rejected alternative is the definition of criminal history, presumably for deciding what information the risk assessment tool evaluates.  Criminal history should be limited to actual convictions.  People are often arrested for crimes they did not commit.  Arrests are often influenced by racial and economic class bias.  An actual conviction should be the only thing included in a person’s criminal history. 

From 2011-2015, over one quarter million people in California were arrested on felonies (this does not include misdemeanor arrests) but did not have charges filed against them.[10]  Prosecutors decline to file charges for a variety of reasons, primarily because there is insufficient evidence to prove the person committed a crime.  To hold an arrest that was rejected for filing against a person in assessing their “risk” is a form of punishment without basis, something the Judicial Council should use its power to forbid.

Similarly, in that same time period, an additional 185,948 people in California charged with felonies had their cases dismissed or won acquittals at trial.  Again, dismissals and acquittals generally reflect a lack of evidence to prove the crime.  The committee should reconsider its decision to allow punishment in the form of higher risk scores for people merely accused of crimes.

Judicial Council Rules

Proposed Rules 4.10 and 4.40 do not provide any definition of the risk categories, low, medium, and high, into which the risk assessment tools will sort people accused of crimes.  They do not impose stringent controls on what the tools will be assessing, what risk they will be assessing for, or even whether they will be transparent and unbiased.  They do not provide stringent procedures to guide local court systems in setting policies concerning “medium” risk people.  Instead, they use vague and general language that allows judges, PAS agents/agencies and local court systems to set their own rules.

Rule 4.10: Proper Use of Risk Assessment Information by the Court

Section 1320.24(a)(1) instructs the Judicial Council to adopt Rules of Court to prescribe the proper use of the pretrial risk assessment information by the court.  This responsibility is an opportunity for the Judicial Council to mitigate some of the harms of risk assessment by establishing reasonable restrictions on their use.  Unfortunately, the rules speak in generalities and do not impose meaningful limits.

Rule 4.10(a) gives a statement of purpose, that the “risk assessment information” should increase public safety and likelihood of return to court, identify “least restrictive” release conditions, and “address any bias” in pretrial decisions. 

Subsection (b)(1) instructs the court to give “significant weight” to the score and to consider the rest of the PAS investigation information.  “Significant weight” is determined by the judge making the decision.[11]  This requirement of “significant weight,” without further definition allows judges to give the score consideration but override it if they find other factors more compelling.  In some jurisdictions already using risk assessment tools, judges over-ride high percentages of release recommendations, while consistently following detain recommendations.[12] 

It is good that the characteristics, needs and interests of the accused are listed as factors to be considered.  However, that information is not required to be considered by the risk assessment tool in generating its recommendation.  The tools are unlikely to consider these characteristics, needs and interests; given the “significant weight” their scores carry, these factors are likely to be disregarded.  The rule could require that risk assessment tools also consider factors about the accused, or it could establish standards with which to weight those factors.

Subsection (b)(2) says that the court and PAS must consider the risk assessment score in context with other information gathered, including information provided by police, victims, attorneys and defendants.  This rule restates Section 1320.10(a) of SB 10, which requires PAS to prepare a report that includes relevant information gathered from police, prosecutors, alleged victims and defendants. 

PAS conducts this investigation pre-arraignment, which means that poor people who need appointed counsel will not yet have their attorney.  Wealthier people with private attorneys on retainer will be able to give PAS mitigating information to increase their chance of a favorable recommendation without the danger of submitting to a personal interview that may result in harmful statements or admissions.  Poor people will not.  Instead, the information gathered will be from police, prosecutors and alleged victims, who all have motive to provide negative information that will decrease the accused person’s chance of release.  The rule does not provide any mechanism to insure fairness by requiring the PAS agent to seek out information about the accused person’s circumstances.  Those with money or strong support networks will get information to the PAS agents, those without will not.  A minimal mitigation to this rule would be to require appointment of council immediately following booking, so that all people will have an opportunity to provide information to help their case. 

Subsection (b)(3) says that the risk score is not determinative, but is relevant to various pretrial detention decisions, including response to alleged violations of conditions and presumptive ineligibility for release[13].  It gives no guidance on how the scores will inform these decisions.  This discretion is preferable to a rule that requires an action be directly tied to a given score, but the complete lack of guidance risks allowing judges to use the scores to justify whatever actions they take.

Subsection (b)(4) says that the court must not use out of date or inaccurate information and investigations.  It does not say that the court cannot use an investigation from a previous arrest, if it is determined that previous investigation is accurate.  The rule does not make clear how the PAS is to make that determination.  Without more precise standards, some PAS agencies may attempt to save money by re-using past investigations and simply asserting that the information in them is accurate.  The rule should forbid using previous assessments.

Subsection (b)(5) advises that the court must consider certain acknowledged flaws[14] of the risk assessment tools.  The rule gives no guidance as to how that consideration should inform decision-making.  Without that guidance, this rule lacks function. 

This rule explicitly allows use of secretive, proprietary tools that do not disclose how they make their recommendations, if judges consider that fact, however they choose to consider it.  Tools that lack fundamental transparency deny due process by making it impossible for the accused person to understand and challenge the reason for the recommendation.  This rule should simply say that tools that do not disclose how they weigh risk factors are not allowed. 

Similarly, this rule tells judges to consider that the risk assessment score is not individualized, but that it gives a statistical likelihood of an outcome based on a group.  In other words, the rule tells judges to understand that they are using a profile.  It does not forbid use of a profile, though profiling violates the principle of individualized justice.  While authorizing non-transparent tools, this rule does not even insure against an explicitly or implicitly racial profile. 

Subsection (b)(5) raises questions about the Judicial Council’s role in approving risk assessment tools that counties may use.[15]  If there has been research raising questions about the tool’s bias, if the tool is not transparent, if it is not properly validated, then Judicial Council should not be approving it for use in the first place.  Instead, this rule assumes that Judicial Council will not filter out tools with these obvious shortcomings, and it will leave judges to decide how they want to address them.  The rule should simply forbid use of tools with these flaws.

Subsection (b)(6) promotes privacy in the otherwise publicly available court file for information gathered by PAS.  However, this information is available to the court and to counsel, including prosecutors.  The Committee has declined to impose a rule that limits use of this information to the pretrial determination only.  This lack of restriction means that prosecutors and judges, with access to the publicly sealed information, may be able to use it against an accused person at other stages of the case, depending on interpretation of Section 1320.9(d).  The material should be sealed from any further view and prevented from being used in other proceedings, as stated previously.

Subsection (c) prescribes “improper uses” of the information and risk scores.  Subsection (c)(1) says courts and the PAS may not use the risk score as the “sole basis” to detain, release or impose conditions on someone, absent specific statutory authorization.  It says that there must be an “individualized evaluation.” 

This rule sounds sensible and appropriate.  The court (or PAS) should make custody determinations based on individualized accounting.  However, the new law and the new rules do not create an adequate structure or standards for such an individualized accounting.  They do not spell out due process requirements or standards for decision-making that are specific enough to constrain judicial discretion.

Instead, the rule should define a fair process and precise standards, like the need for evidence of a specific danger to a specific person or community, that balance safety with the presumption of innocence and imposes a clear, adequate limitations on the weight judges can accord to risk scores.

Subsection (c)(2) makes the important point that the risk of re-offense estimated should be confined to risk during the pretrial period, as opposed to any future offense.  However, because the risk scores are entirely adjustable by the entity that controls them[16], this limitation may not mean much.  If a person has a 20 percent risk of some future offense, but only a 5 percent risk of a new offense during the pretrial period the courts could simply adjust the scoring accordingly to make both of those number correspond to high risk.  Further, this rule illustrates a fundamental mis-understanding of what risk assessment tools estimate.  They cannot estimate risk of re-offense, only of re-arrest.[17] 

This rule should also require a distinction between risk of committing an offense that is a danger as opposed to risk of committing any offense, no matter how minor.

Despite its mandate to define the proper use of risk assessment tools, Rule 4.10 does not create any procedure for accused people to challenge a score generated by the tools.  Without such a procedure, particularly given the vague standards guiding judicial discretion and the tolerance of tools that are not transparent and are possibly biased, the rules do little to instill confidence in fair application of the tools.

Rule 4.40: PAS Standards for Medium Risk People

Section 1320.10 of SB 10 says that people assessed as “medium” risk may either be released or detained depending on standards set by local rules in each county’s court system.  It also says that each local system may create its own exceptions to or categories excluded from pre-arraignment release for medium risk people.

Rules for Release/Detention of “Medium” Risk People

Subsection (a)(2) says that these local rules must maximize release while “reasonably assuring public safety and appearance in court.”  This statement of purpose needs some specific regulations to make it meaningful.

Subsection (b)(2) says PAS must give “significant weight” to the risk assessment score[18], but it does not define how that score is derived and does not say what “significant weight” means.  It does require the PAS to include an explanation of reasons for the release or detain recommendation.  This requirement is important to safeguard against arbitrary decision-making and allow for some review by the court.  It would be helpful for the rule to spell out in more detail how extensive the requirement of a statement of reasons must be.  For example, a checklist type of statement would not be useful and would not improve the thoughtfulness of the decisions.

Subsection (b)(3) lists factors to be considered.  The list is vague.  The circumstances of the crime charged is likely to come only from the arrest report and therefore be extremely one-sided against the accused person.  PAS probably will not conduct independent investigations to get those circumstances.  Any questioning of the accused will implicate 5^th Amendment rights against self-incrimination and 6^th Amendment rights to counsel.  There should be a mechanism, like immediate appointment of counsel, for the accused person to get information about the circumstances of the crime and about the background of the accused person without having to submit to an interview that may harm their defense. 

It is good that PAS must consider the impact of detention on the accused person’s family and community, and not just its impact on victims.  Again, there needs to be some mechanism for the accused person to get this information to the PAS agent without submitting to a potentially compromising interview. 

The rule requires courts to consider criminal history and record of appearance in court as additional factors beyond the risk score.  Many risk assessment tools use these factors in generating their scores.  Therefore, this rule requiring their additional consideration appears to contradict the intent of Rule 4.410(c)(3).  Further, the committee declined to restrict “criminal history” to prior convictions, allowing unproven accusations to be considered.

Subsection (b)(4) says PAS may only incarcerate pre-arraignment if “there is a substantial likelihood that no … conditions of pretrial supervision will reasonably assure public safety or the appearance of the person…”  This standard is raised throughout the new law, but never adequately defined.  Neither the law itself, nor these proposed rules give any guidance or limitation as to how PAS or a judge must make this determination.  Without guidance or limitation, it remains a vague and subjective standard that each judge or PAS agent/agency will apply in their own way.

Subsection (c) sets forth very general rules to guide setting release conditions.  They include using independent judgment, making individualized and not standardized decisions, imposing conditions related only to public safety and return to court, and not making compliance overly difficult.  These are sensible rules, though they remain vague and subject to much interpretation.  The “least restrictive” standard needs to be defined.

The non-exclusive list of possible conditions ranges from non-invasive, like reminders and transportation assistance, to highly restrictive, like electronic monitoring, mandatory medication and transdermal monitoring.  While PAS is told to only impose the “least restrictive” conditions necessary to assure return to court and reduce risk of re-offense, there is no guidance as to how PAS must arrive at that conclusion.  The rules should impose some defined limits on when these severe conditions are allowed, or PAS officers will either resort to formula created by their agency or will exercise their own subjective discretion.  Before requiring a person to take medications or putting them on electronic monitoring[19], there should be significantly more process and opportunity to challenge the order than is provided here.  In fact, there is no opportunity within these rules to object to imposition of a pre-arraignment condition.

Subsection (c)(6) raises a critical point that could be made more decisively.  Pretrial release conditions should not target “rehabilitative objectives related to postconviction supervision.”  The person subject to these conditions has not been convicted of a crime.  Therefore, punitive/rehabilitative conditions are inappropriate.  The rule could make this point more explicitly to clarify that PAS agencies must justify any conditions imposed by their necessity to facilitate return to court and guard against re-arrest only.

Rules for Exclusions on Release

Section 1320.11(a) permits local courts to create their own list of exclusions to release for people assessed as medium risk.  These exclusions create categories of people who simply may not be released at the pre-arraignment stage.  Subsection (d) governs this exercise of power.  Unfortunately, there is no definition of what makes a person “medium” risk.  This classification could include large or small numbers of people, depending on how those who control the scoring choose to define the risk categories.

Subsection (d)(1) adds to the statute’s prohibition on exclusions that apply to everyone in the “medium” risk category, by saying the exclusions cannot affect “nearly all persons.”  This limit does not go far enough in preventing local courts from using this power to incarcerate large percentages of people pretrial.  The exclusions should be targeted to very specific and narrow categories of accused people.

The limitations on what types of exclusions are allowed tend to be vague.  Subsection (d)(2) says exclusions must “uphold the goals of public safety and appearance in court.”  This language could be argued to apply to nearly any exclusion imaginable.  

The rest of the limitations are sensible, if not precisely enough defined.  Status-based exemptions are properly rejected, though the rule should add immigration status as a forbidden exclusion.  Factors weighed by the risk assessment, like missed court dates or prior convictions, should not be subject to blanket exclusion.

Subsection (d)(6) says courts must “consider” if the exclusion increases racial and other bias but gives no direction as to the weight of that consideration.  The rule does not forbid racially discriminatory exclusions or require that an exclusion that disparately impacts one racial group be subject to heightened review or even that it should be disfavored.[20] The rule provides very little guidance or limitation as to the process of creating exclusions.  Local courts are left to create their own procedures that may be as inclusive or exclusive as they want.  Subsection (e)(1) gives minimal direction, none of which amounts to any meaningful check on the courts’ exercise of discretion in creating these exclusions. 

The court “must consult” with PAS and other justice system partners.  The nature of this consultation is not defined.  The consultation may be a matter of showing their partners the changes before implementation.  Those partners only have input if the courts allow it.  Consultation with behavioral health agencies and community organizations is only to the extent the courts deem it “appropriate.”

This rule should be modified to give community and system stakeholders a role in decision-making on these exclusions.

Courts are to review their local rules annually, pursuant to Subsection (e)(2), and examine whether there is a discriminatory impact.  There is no requirement, standard or process to change an exclusion that does have a discriminatory impact.  The report and review have little effect without some remedy for harmful policies.  There is no definition of what data must be analyzed in the annual report.  For the report to be useful, rules must be developed to give courts more guidance on what they must review. 

Missing from the rule is any set of procedures for creating these exclusions.  Each court can decide on its own, as long as they “consult” with the required entities.  The rules should establish uniform procedures that insure exclusions are only created after thoughtful debate amongst a broad variety of stakeholders.

Conclusion

Implementation rules should give precise guidance.  These rules use vague language and give overly broad discretion to judges, courts and PAS agencies to make their own rules.  Principles like “least restrictive conditions” and “public safety” remain undefined.  Even where the rules refer to potential problems with the risk assessment tools or with judicial discretion, their language is permissive and subject to interpretation. 

 The rules should provide limits on discretion, not open-ended justifications for judges and PAS agencies to expand or contract pretrial incarceration at will.

Human Rights Watch opposed passage of SB-10 and these rules do not mitigate the concerns that prompted us to take that position. Human Rights Watch recommends that the Committee extend the time period to develop these rules and re-write them to address the specific problems with SB 10. Please do not hesitate to contact me at raphlij@hrw.org or at (323) 694-5191 if you have any questions.

Signed,

John Raphling
Senior Researcher, Criminal Justice
Human Rights Watch, US Program