Computer simulation of dental professionals as a moral community

Abstract

Current empirical studies of moral behavior of healthcare professionals are almost entirely focused on self-reports, usually collected under the assumption that an ethical disposition characterizes individuals across various contexts. It is well known, however, that individuals adjust their behavior to what they see being done by those in their peer group. That presents a methodological challenge to traditional research within a community of peers because the behavior of each individual is both the result of norms and a contributor to the norms of others. Computer simulations can be used to address this methodological challenge. A Markov replicator model that runs on an Excel spreadsheet was used to investigate a community with four agent types in the dental community: devious practitioners, ethical practitioners who avoid involvement in the poor ethics of others, ethical practitioners who accept it as part of their professional responsibility to challenge colleagues who act unprofessionally, and those who enforce ethical standards. A panel of leaders in the profession independently estimated parameters for the model and criteria for a possible distribution of agent types in the community. The simulation converged on distributions of the agent types that were very similar to the expectations of the panel. The simulation suggests the following characteristics of such moral communities: The structure of such communities is robust across a wide distribution. It appears that reduction in unethical behavior is more sensitive to the way ethical practitioners interact with each other than to sanctions the enforcement community imposes on unethical practitioners, and that large external interventions will be short lived.

Appendix

General findings of a computer simulation of moral behavior of four types of dentists in a moral community.

• 1. Because of the interaction of components, it is almost impossible to determine from inspection what the result of change in a single component will be in the entire system.

• 2. Small changes in sensitive relationships in the system can cause greater disruptions in equilibrium than large changes in slack variables can.

• 3. Systems are often more responsive to relationships among components than to initial proportions of the components.

• 4. Core agents and their interactions establish the system capacity for less central agents. Ethical dentists determine the profession’s tolerance for Devious dentists.

• 5. Complex systems in equilibrium can absorb external interventions and quickly return to the previous steady state.

• 6. In the base model, Good-Active dentists were driven to virtual extinction while Good-Passive dentists retained an essentially undiminished high proportion. This reflects the fact that agents playing similar roles can be substituted functionally for each other to a certain extent and under suitable circumstances.

• 7. Agents that depend on others are highly responsive to changes in the proportion of those agent types. Changes in proportions of Enforcers depend more on changes in proportions of devious Dentists than vice versa.

• 8. Parasite agents depend on the health of others in the system, so they will flourish only if conditions are favorable to those they depend on. They cannot achieve dominance, however, because that would destroy the base on which they depend. Small proportions of Devious dentists and Enforcers are expected to always exist.

The method used here is standard for such research (Dawkins 1989; Maynard Smith 1982; Skyrms 2004, 2010; Sober 2008) and involves multiplying a four-element vector containing the proportion of each agent in the community by itself to produce a four-by-four matrix. This matrix describes the proportion of 16 different types of interactions in the community, on the assumption that interactions are random but proportional to the agents. The four-by-four matrix of proportion of encounters is multiplied by a second four-by-four matrix that describes the effect of each interaction. This multiplication produces a four-element vector of the proportion of each element—adjusted simultaneously for the double effect of impact on other agent types and for the likelihood of encountering them. The “replicator” nature of such models refers to the fact that “successful” encounters increase the proportion of agents following that strategy. The “Markov” nature of such models refers to the fact that the final vector of proportion of each agent type becomes the input vector for the subsequent iteration of the model. Input in the simulation consists of a one-time specification of the initial vector of proportion of agents and a one-time specification of the matrix of the effects of the interactions among agents. The model is depicted below in matrix algebra format, where A = proportion of agents types, O = proportion of opportunities for encounters, and E = effect of encounters.

$$ \left[ {\begin{array}{*{20}l} {{\text{A}}_{1} } \hfill \\ {{\text{A}}_{2} } \hfill \\ {{\text{A}}_{3} } \hfill \\ {{\text{A}}_{4} } \hfill \\ \end{array} } \right]\begin{array}{*{20}l} {{\text{A}}_{1} } \hfill & {{\text{A}}_{2} } \hfill & {{\text{A}}_{3} } \hfill & {{\text{A}}_{4} } \hfill \\ \end{array} = \left[ {\begin{array}{*{20}l} {{\text{O}}_{11} } \hfill & {{\text{O}}_{12} } \hfill & {{\text{O}}_{13} } \hfill & {{\text{O}}_{14} } \hfill \\ {{\text{O}}_{21} } \hfill & {{\text{O}}_{22} } \hfill & {{\text{O}}_{23} } \hfill & {{\text{O}}_{24} } \hfill \\ {{\text{O}}_{31} } \hfill & {{\text{O}}_{32} } \hfill & {{\text{O}}_{33} } \hfill & {{\text{O}}_{34} } \hfill \\ {{\text{O}}_{41} } \hfill & {{\text{O}}_{42} } \hfill & {{\text{O}}_{43} } \hfill & {{\text{O}}_{44} } \hfill \\ \end{array} } \right] \times \left[ {\begin{array}{*{20}l} {{\text{E}}_{11} } \hfill & {{\text{E}}_{12} } \hfill & {{\text{E}}_{13} } \hfill & {{\text{E}}_{14} } \hfill \\ {{\text{E}}_{21} } \hfill & {{\text{E}}_{22} } \hfill & {{\text{E}}_{23} } \hfill & {{\text{E}}_{24} } \hfill \\ {{\text{E}}_{31} } \hfill & {{\text{E}}_{32} } \hfill & {{\text{E}}_{33} } \hfill & {{\text{E}}_{34} } \hfill \\ {{\text{E}}_{41} } \hfill & {{\text{E}}_{42} } \hfill & {{\text{E}}_{43} } \hfill & {{\text{E}}_{44} } \hfill \\ \end{array} } \right] \times \left[ {\begin{array}{*{20}l} 1 \hfill \\ 1 \hfill \\ 1 \hfill \\ 1 \hfill\\ \end{array} } \right] = \left[ {\begin{array}{*{20}l} {{\text{A}}_{1}^{{\prime }} } \hfill \\ {{\text{A}}_{2}^{{\prime }} } \hfill \\ {{\text{A}}_{3}^{{\prime }} } \hfill \\ {{\text{A}}_{4}^{{\prime }} } \hfill \\\end{array} } \right] $$