Automated Decision Making Assessment
Decision-making automation is becoming an increasingly widespread phenomenon, particularly within large organisations, where the adoption of advanced systems is significant. According to recent studies, such systems are used by a considerable portion of large companies (IBPA). Public institutions and organisations consider decision-making automation a consolidated technological opportunity rapidly expanding and redefining operational and strategic dynamics (Alshahrani et al., 2024). The growing diffusion of automation spans several domains, from healthcare to defence, logistics and public services, and the manufacturing and financial sectors. Decision-making automation accelerates the execution of tasks (Langer et al., 2021), but, at the same time, it has the potential to influence the way decisions are made. This may shift the focus from human capabilities and reasoning to technical efficiency logic (Waldman, 2019). Consequently, decision-making automation brings with it crucial questions. How are these processes automated? Which processes can be effectively automated? What are the associated risks? How can the harmful effects be mitigated and governed? While the evidence indicate a growing adoption of Automated Decision-Making Systems, there is a pressing need to understand and assess their social implications fully (Francesco Amato, PhD Thesis Introduction, 2025).
ADMIT: the Assessment Tool
The result of three years of academic research conducted between the University of Naples Federico II and the University of Bristol, this assessment tool for automated decision-making systems (ADMS) is born from a solid scientific path of doctorate. Validated through concrete applications in international industrial contexts, the tool offers an operational and rigorous measurement of the risks, harms and benefits of the social implications related to the adoption and use of ADMS, providing reliable support to decision makers, developers and evaluators.
ADMIT is a research tool developed in the field of social sciences. Its development is strongly interdisciplinary, resulting from the collaboration between academia and industry involving software developers and technology companies. ADMIT is not simply a checklist or a technical protocol: it offers a structured methodological path for assessing the social implications of Automated Decision-Making Systems (ADMS), taking into account the interests, often conflicting, of all stakeholders affected by the system under analysis. The tool consists of two main parts: the risk assessment and the mitigation/support mechanisms, each designed to assess and manage the social implications. It integrates qualitative and quantitative criteria; the latter is operationalised through a seven-level scale that classifies the severity of the risk and the effectiveness of mitigation or protection measures. The development of the research tool began with a conceptual reflection on the nature of Automated Decision-Making Systems (ADMS) to clarify their structural characteristics, operating modes and application contexts. This preliminary phase was essential to delimit the analytical perimeter of the object of study rigorously. Starting from this basis, the work focused on operationalising social implications, understood not as abstract categories but as measurable and assessable effects concerning risks, damages, benefits and forms of inequality that ADMS can produce towards the individuals and groups involved. The transition from theoretical definition to methodological implementation has allowed us to build a tool capable of integrating conceptual analysis with a structured empirical evaluation to identify critical points and possible mitigation strategies. The tool is currently being developed and tested in industrial contexts. During this production phase, the involvement of privileged witnesses and industry experts is essential to validate the scoring criteria and strengthen the tool's structure. This process also includes the development of software applications to support the evaluation workflow and ensure its operational robustness.
The development of the tool began with a conceptual reflection on the nature of Automated Decision-Making Systems (ADMS), in order to clarify their structural characteristics, operating modes and application contexts. This preliminary phase was essential to rigorously delimit the analytical perimeter of the object of study. Starting from this basis, the work then focused on the operationalisation of social implications, understood not as abstract categories, but as measurable and assessable effects in relation to risks, damages, benefits and forms of inequality that ADMS can produce towards the individuals and groups involved. The transition from theoretical definition to methodological implementation allowed us to build a tool capable of integrating conceptual analysis with a structured empirical evaluation, oriented to the identification of critical points and possible mitigation strategies.
The tool is currently being developed and tested in industrial settings. During this production phase, the involvement of key witnesses and industry experts is essential to validate the scoring criteria and strengthen the structure of the tool. This process also includes the development of software applications to support the assessment workflow and ensure its operational robustness.
ADMIT Features
ADMIT is currently structured as an assessment tool in the form of a questionnaire, consisting of two main sections.
The first part is dedicated to the collection of information relating to the subject holding the rights or operational control over the ADMS system being analysed, as well as to the description of the characteristics and functionalities of the system itself. In this section, an accurate operationalisation work has been carried out, with the aim of making detectable, starting from scientific literature and internationally validated frameworks, the actions, configurations and conditions of use typical of automated decision-making systems.
The second part of the tool focuses on mitigation measures. Each item in the questionnaire is associated with a scoring system, through which different values are attributed to the answers. In some cases, the answers provided may determine a reduction in the overall score, improving the final evaluation of the system. The construction of the scores is the result of a research process that integrated theoretical analysis and empirical comparison, through the use of statistical techniques to align the value attributed to each item with that of similar elements already present in other consolidated evaluation tools.
The questionnaire produces two distinct outputs: a risk score and a mitigation score. Both are placed within two 7-level scales, which allow a graduated reading of the severity of the risk associated with the system and the robustness of the compensatory measures envisaged.
The central element of the project is not the questionnaire as such, but the entire methodological path that guides its application, structured to support a reflective, transparent and informed process on the social evaluation of ADMS.
If you wish to learn more or access the tool, you can request contact.
Restricted access – Tool under development as part of a research project
The tool is the result of a research project developed as part of a doctoral thesis, still in progress. It constitutes a significant intellectual asset, the result of theoretical, methodological and applicative work conducted in an academic and interdisciplinary context. For this reason, access to the tool is not public, but can be granted upon motivated request. Sharing is possible within the scope of scientific collaborations, experimental activities or exchanges with institutional and professional subjects, in line with the objectives of the research and with the necessary protections of the work in progress.
