Artificial intelligence (AI) methods and machine learning (ML) are receiving growing attention and have been applied in diverse fields ranging from computational biology to biomedical and medical applications. In the forensic field ML techniques are widely used in the area of Neuroprediction, with applications in assessing the risk of crime recidivism in several contexts, such as prison rehabilitation programs, pretrial risk assessment, and sentencing. However, several ethical concerns have been raised in their massive use and incorporation into the criminal trial regarding their questionable fairness, accountability and transparency.
To overcome these ethical issues, this highly interdisciplinary projects aims at developing models and an (Explainable & Trustworthy) AI-based Decision Support System (DSS) (Virtual Forensic Experts) to guide and support forensic psychiatric evaluations of criminal responsibility and social dangerousness, in order to make them more objective, transparent, and reliable.
Our unique team composition allows us to exploit:
1) Retrospective data available in 320 forensic psychiatric reports made in criminal proceedings and already used as evidence in the trial, provided by 16 forensic psychiatrists from the Northern, Central and Southern Italy.
2) Expert knowledge for all the relevant forensic domains, namely: psychopathology and personality traits of subjects affected by mental illnesses who commit crimes, risk factors of criminal acts, criminal responsibility and social dangerousness, psychiatric epidemiology in subjects who have committed crimes. Expertise on forensic experts' decision making and insanity defense will be covered by a post-doc position.
3) Long-term expertise in AI and Model Checking methods to design AI-based DSS, particularly in the gynecology domain (Melatti, key person and task leader of the EC FP7 PAEON project "Model Driven Computation of Treatments for Infertility Related Endocrinological Diseases").
To date, in the clinical field, AI techniques and machine learning have been mainly applied for diagnosis, prognosis, treatment prediction, and the detection and monitoring of potential biomarkers.
Even more innovative possibilities concern forensic psychiatry, where it is possible to shed new light on different aspects, among which, for example: a) the psychiatric epidemiology in subjects who have committed crimes; b) the risk factors of criminal acts in defendants with mental disorders; c) the predictive elements of accountability, for example in terms of gender, educational attainment, diagnosis (29, 30).
This research project represents the first step in shedding light on forensic evaluators' decisional processes during the insanity assessment.
We will adopt an approach based on the conjoint use of AI KR&R, MC and ML techniques in the forensic evaluations regarding criminal responsibility and social dangerousness. This will allow us to overcome the recent ethical concerns about fairness, accountability, and transparency regarding the massive use of machine learning techniques and their incorporation into the criminal trial. The project purpose is in fact to develop Explainable and Trustworthy AI models which could be transparent with respect to the decision-making process used to reach the final decision and at the same time in line with current expert knowledge and open to change.
The algorithm will be initially developed on the basis of the analysis of Italian forensic reports. In Italy, the legal criteria for evaluating whether a person is not responsible for a crime due to a mental condition are based on a mixed cognitive-volitional assessment, while the evaluation of social dangerousness must consider the probability of crime recurrence in an undefined future (28, 31). Despite insanity definition and the threshold for satisfying its legal criteria tend to vary depending on the jurisdictions, in Western countries, they often rely on the presence of cognitive and/or volitional impairment of the defendant at crime time. A second step of this research project will be the validation of this model in other European Countries (initially the Netherlands and Ireland, with which we already have a research collaboration in the forensic field).
The main purpose of this project is to develop a model to guide and support forensic psychiatric evaluations of criminal responsibility and social dangerousness, in order to make them more objective, transparent, and reliable. This will represent an important advance toward some standardization in an area that is of considerable medical, legal and societal importance, but that regrettably continues to be understudied and will hopefully promote the exchange of ideas and research findings across jurisdictions and disciplines.
REFERENCES (1/2)
1. Gkotsi GM, et al. Neuroscience in forensic psychiatry: From responsibility to dangerousness. Ethical and legal implications of using neuroscience for dangerousness assessments. Int J Law Psychiatry. 2016;46:58-67.
2. Guarnera LA, et al. Field reliability of competency and sanity opinions: a systematic review and metaanalysis. Psychological Assessment. 2017;29(6):795-818.
3. Beckham JC, et al. Decision making and examiner bias in forensic expert recommendations for not guilty by reason of insanity. Law and Human Behavior. 1989;13(1):79-87.
4. Meynen G. Autonomy, Criminal Responsibility, and Competence. J Am Acad Psychiatry Law. 2011;39(231-236).
5. Parmigiani G, et al. Free will, neuroscience, and choice: towards a decisional capacity model for insanity defense evaluations. Riv Psichiatr. 2017;52(1):9-15.
6. Parmigiani G, et al. Translating clinical findings to the legal norm: the Defendant's Insanity Assessment Support Scale (DIASS). Transl Psychiatry. 2019;9(1):278.
7. Janofsky JS, et al. AAPL practice guideline for forensic psychiatric evaluation of defendants raising the insanity defense. Journal of the American Academy of Psychiatry and the Law. 2014;42:S3-S76.
8. Murrie DC, et al. Are forensic experts biased by the side that retained them? Psychol Sci. 2013;24(10):1889-97.
9. Mossiere A, et al. Juror decision making in not criminally responsible on account of mental disorder trials: Effects of defendant gender and mental illness type. Int J Law Psychiatry. 2016;49(Pt A):47-54.
10. Mandarelli G, et al. The factors associated with forensic psychiatrists' decisions in criminal responsibility and social dangerousness evaluations. Int J Law Psychiatry. 2019;66:101503.