1. Overview of the path to change

The administration of the forensic science process falls into the hands of supervisors, local administrators, and prosecution attorneys and police agencies. The process begins with the detection of the evidence of a crime, the collection of the evidence, the selection of the laboratory, the selection of the test, the test of the evidence, the analysis of the evidence, and ends with the communication of the evidence to the decision maker, the judge or jury. The prosecutors’ objective is to maximize convictions. They achieve this objective by determining which arrestees to try and what evidence to bring to the case. Thus, prosecutors have a kind of monopoly power with which they can influence the process of evidence gathering, testing and use. This process leads to bias and bias leads to errors. The current incentive structures and objectives for all parts of the process are aligned, i.e. find evidence of a crime and punish the suspect.

This paper models the ties between the prosecution and forensic scientist as a principal–agent relationship. A principal–agent relationship occurs when a principal hires an agent to carry out work for the principal. An ‘agency problem’ occurs when there are information asymmetries between the principal and the agent, which could lead the actions of the agent to deviate from the expectations of the principal. In the present case, the prosecutor is the principal and the forensic scientist is the agent. When the incentives and objectives of the principal and agent are aligned and there are no information asymmetries, a clear first-best solution to the agency problem exists (Laffont and Martimort, Reference Laffont and Martimort2002: 34). In the present case, the incentives and the objectives of the forensic scientist and the prosecution and police are aligned. As a result, the forensic scientists have an incentive to look for and find evidence to support the prosecution of the suspect. In a world in which forensic science produces truth, all is well.

However, all is not well; the error rate in forensic science is too high.Footnote ¹ Our courts convict people of crimes based on forensic science evidence who are in fact innocent; and we fail to identify and convict perpetrators of crimes who are in fact guilty.Footnote ² A recent National Academy of Sciences study (National Research Council of the National Academies, 2009) found that while advances in forensic science produced tremendous benefits to society,

Additionally, there is an accumulation of DNA exonerations, beginning with that of Gary Dotson in 1989 (Connors et al., Reference Connors, Lundregan, Miller and McEwen1996: 51–52; Garrett and Neufeld, Reference Garrett and Neufeld2009: 4). As of 4 January 2011, the Innocence Project and others have used DNA profiling to exonerate 265 wrongly convicted prisoners in the USA (Innocence Project, 2010). Saks and Koehler (Reference Saks and Koehler2005) authored a widely cited study of the first 86 exonerations conducted by the Innocence Project. Their study showed that forensic science testing errors contributed to the original wrongful conviction in 63% of the cases and false or misleading testimony by forensic scientists played a role in 27% of them (Saks and Koehler, Reference Saks and Koehler2005: 892).Footnote ³ For a review of the relevant literature, see Roger Koppl (Reference Koppl2005), Kelly Pyrek (Reference Pyrek2007) and National Research Council of the National Academies (2009).

A recent series of proposals provides insights on how to minimize these errors including: the basics of ‘competitive self-regulation’ with key changes including rivalrous redundancy of laboratories (Koppl, Reference Koppl2005), forensic experts for defense (Koppl and Cowan, Reference Cowan and Koppl2010) and prosecution alike, ‘sequential unmasking’ (Krane et al., Reference Krane, Ford, Gilder, Inman, Jamieson, Koppl, Kornfield, Risinger, Rudin, Taylor and Thompson2008) effected through a neutral ‘evidence control officer’ (Risinger et al., Reference Risinger, Saks, Thompson and Rosenthal2002) or ‘case manager’ (Krane et al., Reference Krane, Ford, Gilder, Inman, Jamieson, Koppl, Kornfield, Risinger, Rudin, Taylor and Thompson2008), independent laboratory selection (Giannelli, Reference Giannelli1997), a quality-control feedback loop and blind proficiency tests (Whitman and Koppl, Reference Whitman and Koppl2010). (For a more complete review of these proposed reforms, see Koppl, Reference Koppl2005, Reference Koppl2007.) Koppl (Reference Koppl and Lopez2010) addresses cost issues. His study suggests that false-positive errors are so costly that the redundancy of ‘competitive self-regulation’ is likely to produce savings for the criminal justice system as a whole. This paper differs from these contributions by focusing on the organizational structure of individual laboratories, thereby avoiding issues of a cross-jurisdictional nature.Footnote ⁴

These proposed changes are just that – proposed changes. Without providing a path forward, the changes will remain theory in spite of the clear benefits that society would reap. At the heart of the change issue is the relationship between the prosecutor and the forensic science community versus the relationship between the broader society, including the prosecutor (and law enforcement), defense attorney, defendant, criminal justice system, as well as the community that is served by these participants and institutions and the forensic science community. Three examples highlight this failure. In an extreme case, there is the example of Michael Nifong, the District Attorney in the Duke rape case. In that case, three Duke Lacrosse players were accused of rape by an exotic dancer who performed at a party attended by the players. Nifong induced the private DNA laboratory he hired to withhold exculpatory evidence (Zucchino, Reference Zucchino2006). As was subsequently determined, the evidence actually exonerated the players. In an Orange County, California case, the police arrested James Ochoa and held him for 16 months on charges that he had committed robbery and carjacking. In November 2005, Danielle G. Wieland, forensic specialist in the Orange County Sheriff's Department testified in a civil suit for wrongful imprisonment that she had been asked by an Assistant District Attorney to change her ‘no match’ DNA conclusion to indicate that there was a match with Ochoa's DNA. A career criminal was later identified as the perpetrator of the crime (Moxley, Reference Moxley2005, Reference Moxley2008). In the Ochoa case, after trying to influence the crime laboratory, the prosecution ignored their findings in an effort to achieve its goal, i.e. successfully achieve a conviction.

In 1993, Gregory Taylor was convicted of the first-degree murder of Jaquetta Thomas and sentenced to life imprisonment (Swecker and Wolf, Reference Swecker and Wolf2010). On 17 February 2010, a three-judge North Carolina Innocence Inquiry CommissionFootnote ⁵ exonerated Gregory Taylor. The panel found that evidence presented at trial concerning finding Mr Taylor's blood at the crime scene was incomplete. At the time of trial, there was a two-step process to determine whether there were traces of blood at a crime scene. In the first step a screening agent, which is subject to findings of false positives, is used to identify possible blood samples; in the second step, a more complete analysis in the laboratory is conducted. The results from the second test were found to be negative and that information was never passed to the jury. One of the agents involved in this as well as several of the other cases, Special Agent Duane Deaver, who did the serology tests, indicated that it was standard practice to exclude the results of the second test if it was negative (Swecker and Wolf, Reference Swecker and Wolf2010: 7).

I am not using these examples to slam all prosecutors, nor am I saying that all laboratories are willfully dishonest toadies to prosecutors. Rather, I am saying that they are symptomatic of the bigger problem that laboratories tend to view prosecutors or the police, not the public, as their client, which induces bias and, sometimes, cheating to reflect the partial and interested perspective of police and prosecution.

This paper explores redefining the roles of the forensic scientist, the forensic science administration process and the relationships of both with the decision makers in the criminal justice process (the judge or jury). Forensic scientists are agents. The current structure of the criminal justice system makes law enforcement the principal, whereas it should be the public. We first need to change the principal by changing to whom the crime laboratories report. That is not enough. The key to change is the cultural key masters, the supervisors and laboratory directors. We need to provide the right incentives and help them change their identity from law enforcement to science, which is, in part, from the term ‘forensic science’. The bridge that carries us from the old identity to the new identity is built in stages by changing the principal, changing the identity incentives, and changing the compensation incentive. I focus on the ‘middle managers’ at all levels of the forensic science principal–agent structure as the cultural key masters and specifically identify incentive and participation constraints that must be modified in order to make the changes. This paper takes up the challenge of Richard Posner (Reference Posner2010) to apply recent advances in the economics of organizations to institutional arrangements, in order to better understand the institutions and changes needed in those institutions. I lean heavily on recent work in the economics of organizations with special emphasis on work by George Akerlof and Rachel Kranton (Reference Akerlof and Kranton2000, Reference Akerlof and Kranton2002, Reference Akerlof and Kranton2005, Reference Akerlof and Kranton2008) that modifies the worker's utility function to include not only income and effort but also ‘identity’ as an explanatory variable. Identity is the learned relationship that bonds an organization and an individual together and is useful in explaining how and why individuals will perform at a high level when the pay would otherwise be considered insufficient. Additionally, I integrate the traditional behavioral extrinsic motivations assumed by most economists with the recent work of Canice Prendergast (Reference Prendergast2008) by identifying intrinsic motivation as an influence on the willingness or unwillingness of middle managers to change. I use deviations from the ‘identity’ variable to explain why change is difficult to create in forensic science administration and in other organizations where the keepers of the culture, the middle managers, rule.

Finally, even when there is no nefarious intent, this relationship can create bias, which leads to errors. I use Levy and Peart (Reference Levy and Peart2008) to frame the relationship between the prosecutor and the forensic scientist as creating ‘sympathetic bias’. They discuss two ways to compensate an expert: material payment and a desire for approbation (ibid.: 267). Given a class of estimators, which are ‘replicable’ (ibid.: 269), that is without ‘falsification and fabrication’ (ibid.: 265), the expert can choose one that gives least risk of error, lowest standard deviation or they can choose one that gives an answer closer to that sought by the client, but with a higher risk of error. When the statistical distribution is unknown, both answers are compatible with professional standards of acceptable practice.Footnote ⁶ However, when the expert chooses the estimator with the higher risk, she is subject to sympathetic bias (ibid.: 270). Both material incentives and a desire for approbation occur within the context of the prosecutorial–forensic scientist relationship and hence open the door to sympathetic bias and error.

The outline of my argument is as follows. In section 2, I discuss the difficulties in creating change in forensic science administration including size and decentralized nature of the management structure, the complexity of the products/services and the nature of the current alignment of the principal–agent structure in forensic science. In section 3, I examine a modified principal–agent model based on the Akerlof and Kranton utility function. In section 4, I focus on creating the incentives necessary to realign the principal–agent model for forensic science to enable the changes. In section 5, I draw some conclusions and discuss next steps in the path forward.

2. Road blocks to the path forward

3. The principal–agent model and incentives in forensic science

In forensic science there is not an agency problem in the traditional sense. As we shall see, the principal (the prosecution) and the agent (forensic scientist) are aligned in their objective, i.e. to maximize the number of convictions. The issue faced in forensic science is that the forensic scientist is aligned with the wrong principal.

The traditional principal–agent problem arises when the agent has information that the principal does not have and acts differently from what the principal prefers. A principal is an actor or set of actors who must delegate work to an agent whose actions he cannot always observe. The principal's objective function is to maximize profitability at least cost; for the agent, maximize utility at least effort. The agent's utility is a function of compensation and effort. The higher the effort required by the principal, the higher the compensation required by the agent (and vice versa). The objective must be set at a level of effort and associated compensation in order to achieve the profitability targets for the principal. Traditionally, the incentive structure is monetary in nature with either immediate or deferred compensation, i.e. base salary, performance bonuses, deferred stock payment, stock options and other compensation incentives.

Akerlof and Kranton (Reference Akerlof and Kranton2000, Reference Akerlof and Kranton2002, Reference Akerlof and Kranton2005, Reference Akerlof and Kranton2008) modify the traditional worker's utility function to include not only income and effort but also ‘identity’ as an explanatory variable. Identity is derived from the culture of the organization. Identity is the learned relationship that bonds an organization and an individual together and is useful in explaining how and why individuals will execute at a high level when the pay would otherwise be considered insufficient. ‘Identity’ helps the agent frame (Tversky and Kahneman, Reference Tversky, Kahneman, Kahneman and Tversky2000)Footnote ¹⁴ how they make decisions inside an organization and is either a complement or a substitute for income depending on whether the individual identifies himself or herself as an insider or outsider to the organization (Akerlof and Kranton, Reference Akerlof and Kranton2005).

Akerlof and Kranton (Reference Akerlof and Kranton2005) base the identity variable on the psychological theory of personality where individuals internalize the norms, values and beliefs of the group. Internalization of these norms, values and beliefs creates an identity. The norms, values and beliefs typically are based on an idealized form or role model. The role model could be an abstraction, a combination of people or a specific person. Deviations from this ‘ideal’ identity create anxiety. Anxiety causes a reduction in utility. A reduction in utility can be rectified, but only at a cost (Akerlof and Kranton, Reference Akerlof and Kranton2005: 728). Cultures develop an identification process by creating and maintaining the norms, values and beliefs through both subtle and overt mechanisms.

The subtle and overt mechanisms that enforce the identification process include implementing the right ‘tax’ or ‘rhetorical campaign’ (ibid.: 731). More specifically, we can create a penalty structure, an incentive structure or both to mimic the right ‘tax’. For example, a penalty structure could consist of a loss of status, e.g., an individual's job function is reduced to a lower level in the organization or responsibilities removed, or loss of job, loss of bonus, or other pecuniary and non-pecuniary penalties. A positive incentive structure could include recognition in front of peers, increased responsibilities, promotion, as well as other pecuniary and non-pecuniary incentives. The first penalizes deviations from the idealized norms, values and beliefs; the second rewards movements toward the idealized norms, values and beliefs. In the same way that we can influence identity through negative and positive actions, we can influence the culture by creating a ‘rhetorical’ campaign. In managing organizations, this includes developing, communicating and reinforcing, through a rhetorical campaign, a clear set of strategies and associated objectives, a set of shared values, along with the penalty and incentive structures.

What makes an insider an insider? What makes an outsider an outsider? It is not just whether you are a member of the organization (although that may be a necessary condition, it is not a sufficient condition), but rather how you relate to the organization, its culture and its mission and objectives. An insider understands, accepts and reflects the cultural values of the organization. They identify with the organization and can adapt to changes in the organization's culture. They also are able to identify nuances in leadership changes and modify their behavior to match behaviors that are required by the changes. Insiders believe in the norms and goals of the organization. Belief occurs with experiences and experience. Norms are inculcated into the insider by repetition of the norms and the positive reinforcement of pecuniary (particularly bonuses and incremental rewards) and non-pecuniary rewards (recognition, status symbols). Akerlof and Kranton's use of West Point indoctrination techniques to illustrate the transformation from cadet to officer is indicative of the norm-setting process. In non-military organizations, norm setting is often less intensive but normally no less real.

Following Akerlof and Kranton, I modify the standard utility function, which is a function of income and effort, by adding identification utility and define a simple model where insiders (N) and outsiders (O), respectively, have utility functions which can have either a high-action effort (e_jH) or a low-action effort (e_jL). Note that the agency problem occurs when insiders choose a low-action effort (hidden effort), but are compensated for a high-action effort. The high-action effort will tend to increase the possibility of achieving a higher net result, whether profit for the firm, success on the battlefield or in our case, increasing the number of convictions. The low-action effort will tend to achieve a lower net value (Akeroff and Kranton, 2005: 13–14). Identity utility(I_jc) for social category c is added, where social category is depicted as either an insider or outsider. The overall utility function is summarized as U_j(y_j, e_j; c_j, I_jc) = {ln y_j − e_j} + {I_jc − t_jc|e*(c) − e_jc|}, where U_j is the j^th worker's utility, y_j is the worker's income, e_j is the worker's actual effort, e*(c)is the ideal effort and t_c|e*(c) − e_j| is the disutility of diverging from the ideal effort level for category c (ibid.: 14). The identity utility is I_jc = I_j(e _j, c_j; ϵ_j, e*_j), where ϵ_j are j's characteristics and behaviors and e*_c are the ideal norms, values and behaviors for category c_j and e_j are the actual efforts of the j^th individual. The first two parameters (ln y_j, e_j) model the traditional benefit and cost (effort) and the second two parameters model net identity utility. The divergence of the actual effort from the idealized effort causes the anxiety and the loss of identity utility. The factor t_cis used as a scaling factor that scales the difficulty that an individual has in fitting into a group (Akerlof and Kranton, Reference Akerlof and Kranton2002: 1174), e.g., for small t_c even large deviations from the ideal will not cause an individual's utility to decline. Prendergast (Reference Prendergast2008) differentiates between extrinsic motivation and intrinsic motivation. The principal provides extrinsic motivation by the use of monetary and non-monetary incentives to align the agent with the objectives of the principal and to attempt to elicit the level of effort necessary to achieve those objectives. Intrinsic motivation is internally generated by the individual and is not influenced directly by external incentives.Footnote ¹⁵ She indicates ‘that firms, in general, hire agents, who do not share their interests – instead, agents are disproportionately motivated to carry out a subset of what the firm cares about’ (ibid.: 201). Firms focus on hiring individuals who are selected for specialized roles. These specialized roles require specialized skills, with most of the rest of what the firm requires being ignored (ibid.: 201). Prendergast uses as an example welfare agencies hiring individuals who are empathetic to the poor (Prendergast, Reference Prendergast2007) or the Internal Revenue Service (IRS) who are not empathetic with those from whom they are collecting taxes. Some individuals are more willing to change than other individuals simply due to their own internal norms, values and beliefs. I use t_c as a sensitivity-to-changes parameter. An individual with a t_c = 0 indicates that they are indifferent as to deviations from the group norm, regardless of whether |e*(c) − e_j| is large or small, and cannot be influenced by extrinsic motivations, i.e. they are internally motivated. An individual with a 0<t_c < 1, dampens the disutility of deviations from the ideal work effort due to intrinsic motivations, but can be influenced through extrinsic motivations. An individual with a t_c = 1 reduces her utility by the disutility of deviations from the norm without amplifying or dampening the effect. Also, an individual with a t_c > 1 amplifies the effect of deviations from the norm. I speculate that outsiders are more likely to fall in the area 0 ≤ t_c < 1 and insiders in the area t_c ≥ 1.

In forensic science, acquiring an education, working in a laboratory, working with the police and prosecution, testifying before a court of law, accepting membership in and participating in a professional society, becoming an integral part of a forensic science team, recognition of professional work, promotions in the laboratory or in the professional society, and increased remuneration are all mechanisms that create a self-image as a forensic scientist. The longer that an individual is involved in an organization and the more positive the recognition that the individual receives, then the more likely that the individual will perceive of themselves as being an insider. The distance between the self-image and the ideal image and behaviors becomes small |e*(c) − e _j| → 0 and the individual's sensitivity to change t_c becomes large. This leads to a forensic scientist who is dedicated to the current organization, ways of doing work and the existing culture. This dedication as an insider then can become the barrier to change if the insider is in a position of scientific or administrative authority, as is the cases of the supervisors or laboratory directors, i.e. the middle managers of the forensic science hierarchy. I refer to these individuals as the keepers of the culture or the cultural key masters.Footnote ¹⁶ I postulate that as middle managers advance in their profession and identify more closely with the norms, values and objectives of the principal that both |e*(c) − e _j → 0 and t_c become large.

4. Realigning the principal–agent model for forensic science

To enable change we must affect the idealized effort e*(c_j), actual efforts e_j and the sensitivity to change t_c of the cultural key masters. Without their buy-in, changes will trickle in over the next century, much like that of the change from the coroner model to the medical examiner model. For both the small and the larger laboratories, the most important step is to gain the allegiance of the middle managers, the supervisors and laboratory directors. To do this we must provide incentives to them to change their identity from forensic scientist tied to the prosecution to scientist tied to scientific truth. This is the key task at hand. If this fails, the transformation fails and we will continue to live with bias in forensic science and an error rate that is too high. I do note that creating change and sustaining change are not necessarily the same thing. My objective is to create change by providing the appropriate incentives, as well as the institutional realignment necessary to sustain it.

A necessary condition for change in any organization is the existence of senior leadership that establishes the goals and allocates the resources for the change. Once the goals are set and the resources allocated, creating change involves changing both the culture of the organization and the cultural key masters. The cultural key masters are the individuals who control the underlying structure of production and the culture that surrounds that structure of production. Forensic science is a highly fragmented system (National Research Council of the National Academies, 2009: 14), with stand-alone laboratories at the city, county, state and federal levels. Its strength is its distributed nature and its close support of the criminal justice system at the various levels of the system. Its weakness is its distributed nature, since required changes will take extensive efforts to implement and sustain. Leadership is distributed and reports to local prosecutors and police administrators. At best, these leaders bring vision, strategy, direction and a value system to the organization; at worse, they bring chaos.

The workers and scientists produce the product or service. The transmission device between the leaders, prosecution and police and the workers and supervisors is middle management – supervisors and laboratory directors. These are the individuals, who have risen from the ranks of the company or organization by best embodying the culture and mores, and exhibit the ability to ‘get the job done’. They understand how work is done, how to work across internal organizations (Floyd and Wooldridge, Reference Floyd and Wooldridge1997), how to break down barriers and how to build up organizational strength by enhancing their own teams and budgets (Niskanen, Reference Niskanen1975). They know how to survive by changing hats as leadership changes, as leaders change vision, strategy, direction and value systems. They know how to transmit these to the workers and scientists without disrupting the production of products and services. They contain and maintain the ‘organizational memory’,Footnote ¹⁷ and know what works and does not work (within the constraints of the existing organization). Middle managers understand and are experts at optimizing a firm's competences (King et al., Reference King, Fowler and Zeithaml2001). They know that change creates uncertainty at the working level, so they minimize the amount of change transmitted to the workers and supervisors; they act as filters so that the work can continue regardless of changes desired by the leaders of the organization. Balogun and Johnson (Reference Balogun and Johnson2004) borrowing from Weick (Reference Weick1988) refer to this as their role in ‘sensemaking’, literally making sense out of what senior management strategy means. Middle managers have seen leaders come and go; they have seen the chaos created by well-meaning, but ill-informed leaders; they have learned to minimize the chaos by screening the production process from changes of all sorts. They have learned to institute change in a way that minimizes negative impacts on costs and production (Huy, Reference Huy2002). They know that even small changes cause workers to go into the ‘valley of despair’, where all change, good or bad, causes productivity and quality to fall and costs to rise. If change is good, it will eventually cause productivity to rise above the pre-change level; if bad, productivity will stay down permanently and threaten the power of the middle manager.

Middle managers understand the past and present, and know that the leader rarely, if ever, understands the consequences of his or her actions on the day-to-day operations and lives of the workers and scientists. They understand the constraints created by present and past organizations, product and process technologies, information management systems, and human resource systems. They are the masters of maximizing within those constraints. They know that their power lies in the knowledge of how to get the job done within the constraints. They also know that changing the constraints, particularly dramatic changes, changes their own power and their power in relationship to their peers, the workers and scientists and the leaders. The changing of the constraints inevitably diminishes their power and hence threatens their existence (Guth and MacMillian, Reference Guth and MacMillian1986). It is this threat that is at the heart of effectively creating change in an organization.

Change is like death, you must transform from one state of being to another state of being. Individuals must go through the various stages of transformation, from denial, to anger, to bargaining, to moving on and, hopefully, performing (Kubler-Ross, Reference Kubler-Ross1973).

If change is a threat to the existence of the cultural key masters, how does the organization create or respond to market, technological, competitive or governmental changes to the constraints of the organization? The cultural key masters are masters of responding to leaders by creating the illusion of change while controlling and limiting its actual impact. They are quite literally ‘caught in the middle’ between senior leaders and the production process. If they are seen as recalcitrant they are replaced; if they transmit the changes without the appropriate filter, production workers and scientists react by slowing down, becoming less productive, lowering quality or increasing costs, and as a result the middle manager could be replaced. Organizations implicitly understand this dilemma and have attempted to remove layers of middle managers, increase the span of control of middle managers, create quality circles, and create self-managed teams and flat organizations and a host of other approaches without taking on the real issue: changing the ‘cultural key masters’. The cultural key masters are the bottleneck in the change process. If you cannot relieve the bottleneck, the change will fail.

Senior managers cannot be everywhere and know everything; middle managers are necessary to transmit information and coordinate the production process. Since we cannot simply do away with the supervisors or the laboratory directors, then we need to determine ways of turning the middle manager from blockage to facilitator of change. The actions we can take are: (1) change the principal and change the objective; (2) change the identity utility of the agent; and (3) change the compensation incentives structure of the agent.