This thesis work offers an analysis of the phenomenon that for a couple of years now has been distinguishing itself in the continuous and unstoppable technological evolution, i.e. artificial intelligence. This new technology not only places us before ethical and anthropological questions but involves a real paradigm shift at the bottom of which there could be a new idea of humanity. In fact the artificial intelligence and robotics are the biggest technological revolutions of our time, capable of influencing every field of life (education, health, work, entertainment, transport, etc.) and, of course, to have implications for the administrative law. Before such future prospects of technological nature, the administrative law is at a crossroads: in this novelty it can recognize the solution to all problems that concern the public administration, close and refuse completely every possibility of its application to the organization and administrative procedure or, at most, allow a very limited application. The use of systems endowed with artificial intelligence applied to activities of the Public Administration requires us to review (if not even modify) the law institutes, principles and rules of the organization and administrative procedure. In addition, it requires framing correctly the procedural dynamics which arises from it, beyond the relationship between administrative method and the rule of law. In a series of judgments the Council of State recognized the importance of the digitization of the public management for the improvement of the quality of the services rendered to citizens and users, the full correspondence of the algorithm to the standards of efficiency and economy of the administrative action and good trend of the PA, the benefits that you can get in carrying out repetitive activities and without discretion, by means of the exclusion of interference due at the negligence of the official (human being) resulting in greatest impartiality guarantee. At the same time, it specified that the technical formula that disciplines the algorithm is still a general administrative rule, built by man and not by machine, and as such it must respect the general principles of the administrative activity (publicity, transparency, reasonableness, proportionality). Therefore, there are some interrogatives about the usage of the systems gifted of artificial intelligence in the administrative sector (from the computer used as an archive to a computer capable of taking decisions), with particular attention for those that are the concepts, the principles and standards of legitimacy and control, plus that some of their further future applications. One wonders if a system of principles and juridical institutes, thought and built around the official “human being” will be like this easily transferable and at the same time adaptable to the world made of algorithms and smart machines. The Public Administration will become surely very efficient, but at the same time could be deprived completely of that element of humanity that makes the administrative activity, even with all its elements of one- sidedness and imperativeness, acceptable to his recipients. Furthermore, it is necessary to establish if and what space there is for artificial intelligence in a context where the administrative action is characterized from the existence of a right to a good administration, so how is written and detailed in Article 41 of the Charter of fundamental rights of the EU, or, as a synthesis of procedural rights of the citizen, accompanied moreover by a duty to give reasons which does not leave space for certain algorithmic systems. It turns out appropriate to establish adequate methods for integration between the systems endowed with artificial intelligence and man, so that, also in the use of expert system, man keeps providing his cognitive skills that the machine hasn’t got, maintaining the control on the knowledge transferred to him and on the process of its application. On the one hand, the jurist is called to reflect addressed to the present and therefore to the study and interpretation of the current and potential uses of the techniques and of computers tools, as they are accepted by positive law, that also in terms of administrative decisions more and more actually participated and shared by the citizens. On the other side, instead, it will be important to look towards the future of the administration which can't seize unprepared the contemporary legal systems. The investigated scenario, in this thesis work, deals with moreover of the risks highlighted by the doctrine, that is those risks resulting from a block of functionality of the computers systems (crash down), as well as from the insertion of the private powers (such as owners and managers of the computer science technologies, very often subtracted by any form of control by the national State) during the exercise of the public functions, in addition to the possible prejudices that might derive from the storage of the data (in case of cloud computing). A further question pertains to the possible damage that might derive from attacks (also of terrorist matrix) to IT systems of the public administrations. In this moment, the Italian administration has been providing with IT consultants able to deepen the evolutions of artificial intelligence, about exercise of the public functions. Besides, at the legislative level the debate on the impact of the robotics and artificial intelligence inside of the society is still open and in initial phase, waiting to see promulgated the very recent Proposal for a Parliamentary European Regulation and of the Council formulated on April 21, 2021 by EU Commission, which in relation to the method chosen for research, was chosen to concentrate the attention. Despite the doubts and misgivings that arise when you face new themes in one area full of institutes and tested categories, as the administrative right is, the present work means to deal with these new technologies of which is expected to grow the use in the public administration that will pass inevitably from the bureaucracy to the office automation. The present work is structured in the following four chapters. The first chapter deals with offering a synthetic painting about evolution of artificial intelligence, with a space dedicated at the classification and qualification of the emerging technologies in their interconnections with the public administration, in the usage to speed up and simplify its processes and its activities, also thinking about the imputability of the automated decision and correct application of the computer science rules applied to the algorithm. The second chapter, means to analyze the juridical principles whose respect is requested in the use of tools endowed with artificial intelligence. And, after reviewing the fundamental principles of the administrative procedure, and related at the algorithmic decision, you may not consider the international context in which our legal system is inserted. The third chapter intends to observe the adjustment attempts on behalf of the European Union of the phenomenon, analysing in the detail the Parliament 's Proposal for a European Regulation and Council that establishes rules harmonized on artificial intelligence and modifies some legislative deeds of the Union. Finally, the fourth chapter is dedicated to the future prospects. It will try to examine the spaces, the limits and the risks that arise with the introduction of the algorithmic instrument in the administrative system. In particular it will try to evaluate a concrete usage of the AI systems in administrative method as possible risk of a minor boundary line between machine and person in the figure of a robot as the mouth of the law. The digitization process that is passing through the actual society has a constant impact of growth about the administration and the way in which it operates. This process not only involves the dematerialization of the public activity and relationships between administrative and private authorities, with the aim of simplification and modernization, but brings with it also the use of tools and digital technologies that modify the power and the ways of its exercise. The equipped systems with AI, however, beyond that provide to the public authorities significant gains in terms of efficiency and knowledge, thanks to powerful analytics tools of the data and prediction, they carry with them some risks, on which the attention of the Italian doctrine and beyond has been focused for a long time. Currently the risks are multiple for privacy, security risks of the citizens, for the fundamental rights and for some democrats principles. Also, there is no certainty that the systems equipped with AI are compatible with the fundamental principles of the action administrative, such as transparency, motivation, participation and impartiality due to their opacity and difficulty that they present in understanding them. It seems legitimate to wonder if a nucleus of institutes and fundamental principles built around the human official may be applicable to a PA called to decide through algorithms and, therefore, through the use of “intelligent” machines. Currently different degrees of advancement are recorded in the use of AI systems with reference to central administrations such as INPS and Agenzia dell’entrate, in fact, these institutions are united by a very advanced digitization process and by experimentation with established chatbots and AI systems, both in anti-fraud operation, in the fight against fiscal tax evasion. The ministry of Justice, on the other hand, is characterized by a slow and uneven digitization process in the different sectors of the justice, with evident relapses on possible use of systems equipped with AI. However, one of the causes of the delay in experiments about the use of new technologies for the administration of justice is its decentralization that allows it to operate only on limited samples of data. The administrations are experimenting AI tools outside a normative framework that clearly defines their use within the procedure and the role of the decision human maker. Besides, it seems to be absent an internal policy aimed at equipping administration to the human surveillance, starting from the training staff, which needs to be able to understand logics and the operation modes of the AI systems, thus promoting engagement of the private individuals, whose management skills of the systems result to be strengthen. The latter question is intended to acquire a major relevance in the next future, as proposed in the regulation of the European Commission, in line with what already established by the GDPR, identifies in the “human surveillance” one of requirements required for the use of the AI systems. On the relationship between the tools equipped with AI and the administrative process, according to authoritative doctrine there is the possibility of a complete replacing of the judge with a machine, though they are in favor for the usage of it as an assistant to the human decision-maker. The judge needs to keep full autonomy of judgment and must not base his decision exclusively on directions coming from a computer. Therefore, there is an orientation that pushes for the usage of the machine as an aid to the human judge, assigning it the role of a sort of general robotic lawyer, that will have the task of providing the human judge technical evaluations, the reconstruction of the facts, as well as preparing drafts of the measures and for the treatment of serial, repetitive cases. So the human judge, in total autonomy, will assess this decision or disregard it but giving reasons (Ruffolo, 2021). Therefore, a system is proposed that preserve the primacy of the human rights decision and the absolute freedom of judgment of the human judge, placing the machine side by side with a servant function, which will provide a draft sentence that the human judge will be free to disregard, but decisively justified. For some years administrative jurisprudence has faced this theme by means of the application of the general principles and in particular through the so called algorithmic reasonableness. Some judgments of the TAR have been significant in which the administrative judge has recognized lack of preliminary investigation, lack of motivation and misrepresentation of the facts as at the basis of the analysis there were textual correlations between inconsistent data with the scientific assessment, which combined with other indexes have led to distort the final result. Also, from hermeneutics adopted by some rulings of the State Council, emerges as the administrative automated decision requires the judge to assess the fairness of the IT process in all its components: from its construction, the insertion of the data, to their validity, to their management. Indeed, in the event of a decision founded on an algorithm, a declination is ensured strengthened by the principle of transparency, understood as full knowability of the rule expressed in a different language from that juridical one. Furthermore, since it is a decision made by a public administration, also highlights the art. 41 of the Charter of Nice, and the right to a “good administration”, where it is claimed that, when public administration means to make a decision that can have adverse effects on a person, it must hear him/her before acting, to allow him/her the access to its archives and documents, as well as to “give its reasons for its decisions”. The judge’s decision must be justified moreover, so that the parties understand the reasons but also to allow to be contested before a superior judge. Any decision implies a choice and any choice presupposes a selection of values and purposes. If this applies to human being, a doubt arises for the eventual algorithm that it seems decide without choosing, but applying a series of calculations for the solution of the problems. However, neither the automatic decision really is neutral as to operate before and for the operation of the machine there is always the man, who acts on the basis of options of value and/or interest. This aspect is of is crucial importance for the analysis of the theme as the charm of the algorithm resides on the belief that the perfect measure is that free from errors and that, not to commit mistakes, no may be human; that the car is not influenced, upset, polluted by emotions or matters of conscience and that it is not susceptible to corruption; that it makes sure efficient results, since it reduces response times and so on prevents slopes; that offers effective results because more suitable to the care and to the satisfaction of the pursued goals, guaranteeing impartial solutions. But the decision qualitatively ideal is not the “perfect” one, but rather the “right” one, where the predetermined rules must be respected and adapt to the subjective situation of the care. Furthermore, efficiency is not a prerogative of the machine, also the man can be efficient when his activity is organized and with suitable resources. Finally, the impartiality of the machine if apparently implemented by the adopted decision, may not be, during the identification of the processing criteria of the algorithm. Those which emerge from the survey, in fact, are administrations that want to experiment innovative technologies to improve quality and effectiveness of their own activity. The administrations, in other words, try to use the AI systems because these allow to innovate radically the ways of carrying out the activity of which they are responsible and the expected results, regardless functional specificity of each administration. Therefore, it is necessary an effective legislative solution, such as a code of technological conduct, in whose context the whole matter is disciplined and in which could merge both technical themes and more juridical issues connected to the technologies that there are in the public administration, to administrative organization that believes to be preferable to the administrative procedure, to the imputation and liability model. Then we must consider that currently, there are no AI systems capable of replacing a lawyer completely. Systems which, however, are able to arrange – with the slightest cooperation of a human being – legal simple or repetitive acts, extract norms, precedent judgments, maxims, arguments, or to analyse documents concerning states of reality and understand them at the equal to an human operator for the limited purposes to which their knowledge is useful. If it is certain that in the next future we will assist at the development of AI systems more and more autonomous and able to take decisions complex in several fields of applications; it turns out essential to understand if ethics problems related to such systems must counter this technological development, or if there is a way to implement a useful and responsible technological development that may bring benefits for humans. In conclusion, a very important role in the AI systems is played by transparency, in fact, the same needs to be secured by the right of access to the algorithm, from its “openness ” understood as the possibility of making manifest and understandable to the man the way to operate of the autonomous systems, the reasons that they have brought to certain choices, and the concerning risks, with the aim of maintaining control of man on the machine , communicate and interact with AI systems and have knowledge of their actions. In view of what has been analyzed in this work it is possibile to try to draw a line of continuity with those that will be future researches on this theme, it will be interesting to understand what the future will be of the central figures in the administrative process (for example the person in charge of the procedure). At the same time, it will be interesting to see the effects of the application of a regulation on artificial intelligence for purposes of the European data governance and control on market for these smart systems.
Il presente lavoro di tesi offre un’analisi del fenomeno che da un paio di anni a questa parte si sta contraddistinguendo nella continua e infrenabile evoluzione tecnologica, ovvero l’intelligenza artificiale, questa nuova tecnologia non solo ci pone davanti a interrogativi di tipo etico e antropologico ma comporta un vero e proprio cambio di paradigma al fondo del quale potrebbe esserci una nuova idea di umanità. L’intelligenza artificiale e la robotica si candidano difatti ad essere le più grandi rivoluzioni tecnologiche del nostro tempo, capaci di influenzare ogni ambito della vita (educazione, salute, lavoro, intrattenimento, trasporti, ecc.) e, certamente, di avere implicazioni per il diritto amministrativo. Dinnanzi a tali prospettive future di carattere tecnologico, il diritto amministrativo si trova a un bivio: può riconoscere in questa novità la soluzione a tutti i problemi che riguardano la pubblica amministrazione, chiudersi e rifiutare completamente ogni possibilità di una sua applicazione all’organizzazione e al procedimento amministrativo o, al massimo, consentirne un’applicazione molto limitata. L’uso dei sistemi dotati di intelligenza artificiale applicati alle attività della Pubblica Amministrazione impone di riesaminare (se non addirittura di modificare) gli istituti giuridici, principi e regole dell’organizzazione e del procedimento amministrativo. In più, impone di inquadrare correttamente la dinamica procedimentale che ne scaturisce, oltre al rapporto fra procedimento amministrativo e il principio di legalità. Il Consiglio di Stato in una serie di sentenze ha riconosciuto l’importanza della digitalizzazione della pubblica amministrazione per il miglioramento della qualità dei servizi resi ai cittadini e agli utenti, la piena rispondenza dell’algoritmo ai canoni di efficienza ed economicità dell’azione amministrativa e del buon andamento della P.A., i benefici che si possono ottenere nello svolgimento di attività ripetitive e prive di discrezionalità, mediante l’esclusione di interferenze dovute alla negligenza del funzionario (essere umano) con conseguente maggior garanzia di imparzialità. Allo stesso tempo, ha precisato che la formula tecnica che disciplina l’algoritmo è pur sempre una regola amministrativa generale, costruita dall’uomo e non dalla macchina, e come tale deve rispettare i principi generali dell’attività amministrativa (pubblicità, trasparenza, ragionevolezza, proporzionalità). Pertanto, ci si pongono degli interrogativi sull’utilizzo dei sistemi dotati dell’intelligenza artificiale nel settore amministrativo (dal computer utilizzato come un archivio ad un computer capace di prendere decisioni), con particolare attenzione a quelli che sono i concetti, i principi e gli standard di legittimità e controllo, oltre che alcune delle loro ulteriori future applicazioni. Ci si chiede se un sistema di principi e istituti giuridici, pensato e costruito attorno al funzionario “essere umano” sarà così facilmente trasferibile e allo stesso tempo adattabile al mondo fatto di algoritmi e di macchine intelligenti? La Pubblica Amministrazione diventerà sicuramente efficientissima, ma al contempo potrebbe essere privata completamente di quell’elemento di umanità che rende l’attività amministrativa, pur con tutti i suoi elementi di unilateralità e imperatività, accettabile ai suoi destinatari. Inoltre, occorre stabilire se e quale spazio vi sia per l’intelligenza artificiale in un contesto in cui l’azione amministrativa è caratterizzata dall’esistenza di un diritto ad una buona amministrazione, così com’è scritto e dettagliato nell’articolo 41 della Carta dei diritti fondamentali dell’UE, ovvero, come sintesi di diritti procedimentali del cittadino, accompagnati per di più da un dovere di motivazione che non lascia facile spazio a determinati sistemi algoritmici. Risulta opportuno stabilire adeguate modalità per l’integrazione tra i sistemi dotati di intelligenza artificiale e l’uomo, affinché, anche nei casi di utilizzo di sistemi esperti, l’uomo continui a fornire le proprie competenze cognitive di cui la macchina è priva, mantenendo il controllo sulla conoscenza trasferitagli e sul processo della sua applicazione. Da un lato il giurista è chiamato a una riflessione rivolta al presente e quindi allo studio e all’interpretazione delle utilizzazioni attuali e potenziali delle tecniche e degli strumenti informatici, così come sono accolte dal diritto positivo, ciò anche in funzione di decisioni amministrative sempre più effettivamente partecipate e condivise dai cittadini. Dall’altro lato, invece, sarà importante proiettarsi con lo sguardo verso il futuro dell’amministrazione che non può cogliere impreparati gli ordinamenti contemporanei. Lo scenario indagato, nel lavoro di tesi, tratta inoltre dei rischi evidenziati dalla dottrina, ovvero quei rischi derivanti da un blocco di funzionalità dei sistemi informatici (crash down), nonché dall’inserimento dei poteri privati (quali proprietari e gestori delle tecnologie informatiche, molto spesso sottratti da ogni forma di controllo da parte dello Stato nazionale) nell’esercizio delle pubbliche funzioni, oltre ai possibili pregiudizi che potrebbero derivare dallo stoccaggio dei dati (in caso di cloud computing). Un’ulteriore questione attiene agli eventuali danni che potrebbero derivare da attacchi (anche di matrice terroristica) ai sistemi informatici delle pubbliche amministrazioni. In questo momento, l’amministrazione italiana si sta dotando di consulenti informatici capaci di approfondire le evoluzioni dell’intelligenza artificiale, sull’esercizio delle pubbliche funzioni. Inoltre, a livello legislativo il dibattito sull’impatto della robotica e dell’intelligenza artificiale all’interno della società resta tuttora aperto e in fase iniziale, in attesa di vedere promulgata la recentissima Proposta di Regolamento del Parlamento europeo e del Consiglio formulata il 21 aprile 2021 dalla Commissione UE, cui in relazione al metodo prescelto per la ricerca, si è scelto di concentrare l’attenzione. Il presente lavoro intende così occuparsi, nonostante i dubbi e le perplessità che insorgono quando si affrontano tematiche nuove in un ambito pieno di istituti e categorie collaudate, qual è il diritto amministrativo, di queste nuove tecnologie delle quali è prevedibile un crescente utilizzo da parte della pubblica amministrazione che passerà inevitabilmente dalla burocrazia alla burotica. Ciò posto, il presente lavoro è strutturato nei seguenti quattro capitoli. Il primo capitolo si occupa di offrire un sintetico quadro sull’evoluzione dell’intelligenza artificiale, con uno spazio dedicato alla classificazione e qualificazione delle tecnologie emergenti nelle loro interconnessioni con la pubblica amministrazione, nel loro utilizzo per velocizzare e semplificare i suoi processi e le sue attività, fino a porre uno sguardo all’imputabilità della decisione automatizzata e la corretta applicazione delle regole informatiche applicate all’algoritmo. Con il secondo capitolo, si intende analizzare i principi giuridici di cui si richiede il rispetto nell’utilizzo di strumenti dotati di intelligenza artificiale. E, dopo aver passato in rassegna i principi fondamentali del procedimento amministrativo, e in relazione alla decisione algoritmica, non si può non considerare il contesto internazionale nel quale il nostro ordinamento si inserisce. Proprio il terzo capitolo si propone di osservare i tentativi di regolazione europea del fenomeno, analizzando nel dettaglio la Proposta di Regolamento del Parlamento Europeo e del Consiglio che stabilisce regole armonizzate sull’intelligenza artificiale e modifica alcuni atti legislativi dell’Unione. Infine, il quarto capitolo è dedicato alle prospettive future, si proverà ad esaminare gli spazi, i limiti ed i rischi che sorgono con l’introduzione dello strumento algoritmico nel sistema amministrativo. In particolare si proverà a valutare un concreto utilizzo dei sistemi di IA nel procedimento amministrativo, il possibile rischio di un minore linea di confine tra macchina e persona nella figura di un robot quale bocca della legge. Il processo di digitalizzazione che sta attraversando la società odierna ha un impatto in costante crescita sull’amministrazione e sul modo in cui essa opera. Tale processo non solo comporta la dematerializzazione dell’attività pubblica e dei rapporti tra autorità amministrative e privati, con intenti di semplificazione e modernizzazione, ma porta con sé anche l’impiego di strumenti e tecnologie digitali che modificano il potere e i modi del suo esercizio. I sistemi dotati di IA, però, oltre che fornire alle autorità pubbliche significativi vantaggi in termini di efficienza e conoscenza, grazie a strumenti potenti di analisi dei dati e di predizione, portano con sé dei rischi, su cui da tempo si è focalizzata l’attenzione della dottrina italiana e non solo. Allo stato attuale i rischi sono molteplici ovvero rischi per la privacy, per la sicurezza dei cittadini, per i diritti fondamentali e per gli stessi principi democratici. Inoltre, non vi è la certezza che i sistemi dotati di IA siano compatibili con i principi cardine dell’azione amministrativa, quali la trasparenza, la motivazione, la partecipazione e l’imparzialità a causa della loro opacità e delle difficoltà che essi presentano nel comprenderli. Appare lecito domandarsi se un nucleo di istituti e principi fondamentali costruiti attorno al funzionario umano possa essere applicabile ad una P.A. chiamata a prendere decisione sulla base di algoritmi e, quindi, tramite l’utilizzo di macchine “intelligenti”. Ad oggi si registrano diversi gradi di avanzamento nell’impiego di sistemi di IA con riferimento alle amministrazioni centrali quali INPS e Agenzia delle entrate, infatti, le due istituzioni sono accomunate da un processo di digitalizzazione molto avanzato e dalla sperimentazione di chatbot e sistemi di IA consolidati, sia in funzione antifrode, che di lotta all’evasione fiscale. Il ministero della Giustizia, invece, è contraddistinto da un processo di digitalizzazione lento e disomogeneo nei diversi settori della giustizia, con evidenti ricadute sul possibile utilizzo di sistemi dotati di IA. Tuttavia, una delle cause del ritardo nelle sperimentazioni sull’uso di nuove tecnologie per l’amministrazione della giustizia è la sua decentralizzazione che gli consente di operare soltanto su campioni limitati di dati. Le amministrazioni stanno sperimentando strumenti di IA al di fuori di un quadro normativo che definisca in modo chiaro il loro impiego all’interno del procedimento e il ruolo del decisore umano. Inoltre, sembra essere assente una politica interna volta ad attrezzare l’amministrazione alla sorveglianza umana, a partire dalla formazione del personale, che deve essere in grado di comprendere la logica e i modi di funzionamento dei sistemi di IA, favorendo in questo modo il coinvolgimento dei privati, le cui capacità di gestione dei sistemi risultano essere rafforzate. Quest’ultima questione è destinata ad acquisire una maggiore rilevanza nel prossimo futuro, la già richiamata proposta di Regolamento della Commissione europea, in linea con quanto già stabilito dal GDPR, individua nella “sorveglianza umana” uno dei requisiti obbligatori per l’utilizzo dei sistemi di IA. Sul fronte del rapporto tra i strumenti dotati di IA e il processo amministrativo, secondo autorevole dottrina non vi è la possibilità di una completa sostituzione del giudice con una macchina, sebbene gli stessi favoriscono per l’utilizzo della stessa quale ausiliare del decisore umano. Il giudice deve mantenere la piena autonomia di giudizio e non deve basare la propria decisione esclusivamente sulle indicazioni provenienti da un computer. Pertanto, vi è un orientamento che spinge per l’utilizzo della macchina quale ausilio del giudice umano, assegnandole la veste di una sorta di Avvocato Generale robotico, che avrà il compito di fornire al giudice umano le valutazioni tecniche, la ricostruzione dei fatti, così come la predisposizione di bozze dei provvedimenti e per la trattazione di cause seriali, repetitive e interamente documentali. Spetterà, dunque, al giudice umano, in totale autonomia, censire tale decisione o disattenderla ma motivando (Ruffolo, 2021). Si propone, dunque, un sistema che salvaguardi il primato della decisione umana e la assoluta libertà di giudizio del giudice umano, affiancandogli la macchina con una funzione servente, la quale fornirà un progetto di sentenza che il giudice umano sarà libero di disattendere, ma con decisione motivata. Da alcuni anni la giurisprudenza amministrativa affronta questa tematica partendo dall’applicazione dei principi generali e in particolare di quella che può essere chiamata la ragionevolezza algoritmica. Significative sono state alcune sentenze del TAR in cui il giudice amministrativo ha ravvisato difetto di istruttoria, carenza di motivazione e travisamento dei fatti in quanto alla base dell’analisi vi erano correlazioni testuali tra dati non coerenti con la valutazione scientifica, che combinate ad altri indici hanno portato a sfalsare il risultato finale. Inoltre, dall’ermeneutica adottata da alcune pronunce del Consiglio di Stato, emerge come la decisione amministrativa automatizzata impone al giudice di valutare in primo luogo la correttezza del processo informatico in tutte le sue componenti: dalla sua costruzione, all’inserimento dei dati, alla loro validità, alla loro gestione. Infatti, in caso di decisione fondata su un algoritmo, è assicurata una declinazione rafforzata del principio di trasparenza, intesa come piena conoscibilità della regola espressa in un linguaggio differente da quello giuridico. Inoltre, trattandosi di una decisione presa proprio da una pubblica amministrazione, viene in rilievo anche l’art. 41 della Carta di Nizza, ed il diritto ad una “good administration”, laddove si afferma che, quando la pubblica amministrazione intende adottare una decisione che può avere effetti avversi su di una persona, essa ha l’obbligo di sentirla prima di agire, di consentirle l’accesso ai suoi archivi e documenti, nonché di “motivare le proprie decisioni”. La decisione del giudice va inoltre motivata, perché le parti ne capiscano le ragioni ma anche per consentire di contestarla dinanzi a un giudice superiore. Qualsiasi decisione implica una scelta e qualsiasi scelta presuppone una selezione di valori e finalità. Se questo vale per l’essere umano, un dubbio sorge per l’eventuale algoritmo che sembra decidere senza scegliere, bensì applicando una serie di calcoli per la soluzione dei problemi. Tuttavia, neanche la decisione automatica è davvero neutrale poiché ad operare prima e per il funzionamento della macchina vi è sempre l’uomo, che agisce in base ad opzioni di valore e/o di interesse. Questo aspetto è di fondamentale importanza ai fini dell’analisi della tematica in quanto il fascino dell’algoritmo risiede sulla convinzione che il provvedimento perfetto sia quello privo di errori e che, per non commettere errori, non si possa essere umani; che la macchina non sia influenzata, turbata, inquinata da emozioni o questioni di coscienza e che non sia suscettibile di corruzione; che essa assicuri risultati efficienti, poiché riduce i tempi di risposta e così previene le pendenze; che offra esiti efficaci, poiché maggiormente idonei alla cura e alla soddisfazione degli obiettivi perseguiti, garantendo soluzioni imparziali. Ma la decisione qualitativamente ideale non è quella “perfetta”, bensì quella “giusta”, ovvero quella che rispetti le regole predeterminate e si adegui alla situazione soggettiva di cura. Inoltre, l’efficienza non è una prerogativa della macchina, anche l’uomo sa e può essere efficiente quando la sua attività è organizzata e assistita da risorse idonee. Infine, l’imparzialità della macchina se apparentemente attuata a valle della decisione adottata, può non esserlo a monte, in sede di individuazione dei criteri di elaborazione dell’algoritmo. Quelle che emergono dall’indagine, infatti, sono amministrazioni che vogliono sperimentare tecnologie innovative per migliorare la qualità e l’efficacia delle proprie attività. Le amministrazioni, in altri termini, si orientano verso i sistemi di IA perché questi ultimi permettono di innovare radicalmente i modi di svolgimento delle attività di cui sono responsabili e i risultati attesi, indipendentemente dalle specificità funzionali di ciascuna amministrazione. Pertanto, ritengo che sia strettamente necessaria una soluzione normativa efficace, quale ad esempio un codice di condotta tecnologica, nel cui contesto disciplinare l’intera materia e nel quale potrebbero confluire sia temi di natura squisitamente tecnica sia questioni più prettamente giuridiche legate all’uso delle tecnologie emergenti da parte della pubblica amministrazione, alla organizzazione amministrativa che si ritiene preferibile, al procedimento amministrativo, al modello di imputazione e di responsabilità. Si consideri poi che allo stato attuale, non si dispone di sistemi di IA in grado di sostituire un giurista completamente. Sistemi che, tuttavia, sono in grado di predisporre – con la minima collaborazione di un essere umano – atti giuridici semplici o ripetitivi, estrarre norme, precedenti giudiziali, massime, argomentazioni, oppure di analizzare documenti riguardanti stati di realtà e comprenderli alla pari di un operatore umano per i limitati fini ai quali la loro conoscenza è utile. Potendo dare quasi per certo che nel prossimo futuro si assisterà allo sviluppo di sistemi di IA sempre più autonomi e in grado di prendere decisioni complesse in diversi campi applicativi; risulta indispensabile comprendere se le problematiche etiche relative a tali sistemi debbano contrastare tale sviluppo tecnologico, oppure se ci sia un modo per attuare uno sviluppo tecnologico utile e responsabile che possa portare vantaggi per l’uomo. In conclusione, un ruolo molto importante nei sistemi di IA lo gioca la trasparenza, infatti, la stessa deve essere assicurata dal diritto di accesso all’algoritmo, dalla sua “apertura” intesa come la possibilità di rendere manifesta e comprensibile all’uomo la modalità di operare dei sistemi autonomi, le motivazioni che hanno portato a certe scelte, ed i relativi rischi, con l’obiettivo di mantenere il controllo dell’uomo sulla macchina, comunicare ed interagire con i sistemi di IA ed avere conoscenza del perché del loro operato. Sulla scia del presente lavoro si proverà a tracciare una linea di continuità con quelle che saranno le future ricerche su questa tematica, sarà interessante capire quale sarà il futuro delle figure centrali del procedimento amministrativo (ad esempio il responsabile del procedimento). Allo stesso tempo, sarà interessante vedere gli effetti dell’applicazione di un regolamento sull’intelligenza artificiale ai fini della data governance europea e del controllo sul mercato di questi sistemi intelligenti.
INTELLIGENZA ARTIFICIALE E PUBBLICA AMMINISTRAZIONE: RIFLESSIONI, SFIDE E OPPORTUNITÀ
Puzzanghera, Alessandro
2023
Abstract
This thesis work offers an analysis of the phenomenon that for a couple of years now has been distinguishing itself in the continuous and unstoppable technological evolution, i.e. artificial intelligence. This new technology not only places us before ethical and anthropological questions but involves a real paradigm shift at the bottom of which there could be a new idea of humanity. In fact the artificial intelligence and robotics are the biggest technological revolutions of our time, capable of influencing every field of life (education, health, work, entertainment, transport, etc.) and, of course, to have implications for the administrative law. Before such future prospects of technological nature, the administrative law is at a crossroads: in this novelty it can recognize the solution to all problems that concern the public administration, close and refuse completely every possibility of its application to the organization and administrative procedure or, at most, allow a very limited application. The use of systems endowed with artificial intelligence applied to activities of the Public Administration requires us to review (if not even modify) the law institutes, principles and rules of the organization and administrative procedure. In addition, it requires framing correctly the procedural dynamics which arises from it, beyond the relationship between administrative method and the rule of law. In a series of judgments the Council of State recognized the importance of the digitization of the public management for the improvement of the quality of the services rendered to citizens and users, the full correspondence of the algorithm to the standards of efficiency and economy of the administrative action and good trend of the PA, the benefits that you can get in carrying out repetitive activities and without discretion, by means of the exclusion of interference due at the negligence of the official (human being) resulting in greatest impartiality guarantee. At the same time, it specified that the technical formula that disciplines the algorithm is still a general administrative rule, built by man and not by machine, and as such it must respect the general principles of the administrative activity (publicity, transparency, reasonableness, proportionality). Therefore, there are some interrogatives about the usage of the systems gifted of artificial intelligence in the administrative sector (from the computer used as an archive to a computer capable of taking decisions), with particular attention for those that are the concepts, the principles and standards of legitimacy and control, plus that some of their further future applications. One wonders if a system of principles and juridical institutes, thought and built around the official “human being” will be like this easily transferable and at the same time adaptable to the world made of algorithms and smart machines. The Public Administration will become surely very efficient, but at the same time could be deprived completely of that element of humanity that makes the administrative activity, even with all its elements of one- sidedness and imperativeness, acceptable to his recipients. Furthermore, it is necessary to establish if and what space there is for artificial intelligence in a context where the administrative action is characterized from the existence of a right to a good administration, so how is written and detailed in Article 41 of the Charter of fundamental rights of the EU, or, as a synthesis of procedural rights of the citizen, accompanied moreover by a duty to give reasons which does not leave space for certain algorithmic systems. It turns out appropriate to establish adequate methods for integration between the systems endowed with artificial intelligence and man, so that, also in the use of expert system, man keeps providing his cognitive skills that the machine hasn’t got, maintaining the control on the knowledge transferred to him and on the process of its application. On the one hand, the jurist is called to reflect addressed to the present and therefore to the study and interpretation of the current and potential uses of the techniques and of computers tools, as they are accepted by positive law, that also in terms of administrative decisions more and more actually participated and shared by the citizens. On the other side, instead, it will be important to look towards the future of the administration which can't seize unprepared the contemporary legal systems. The investigated scenario, in this thesis work, deals with moreover of the risks highlighted by the doctrine, that is those risks resulting from a block of functionality of the computers systems (crash down), as well as from the insertion of the private powers (such as owners and managers of the computer science technologies, very often subtracted by any form of control by the national State) during the exercise of the public functions, in addition to the possible prejudices that might derive from the storage of the data (in case of cloud computing). A further question pertains to the possible damage that might derive from attacks (also of terrorist matrix) to IT systems of the public administrations. In this moment, the Italian administration has been providing with IT consultants able to deepen the evolutions of artificial intelligence, about exercise of the public functions. Besides, at the legislative level the debate on the impact of the robotics and artificial intelligence inside of the society is still open and in initial phase, waiting to see promulgated the very recent Proposal for a Parliamentary European Regulation and of the Council formulated on April 21, 2021 by EU Commission, which in relation to the method chosen for research, was chosen to concentrate the attention. Despite the doubts and misgivings that arise when you face new themes in one area full of institutes and tested categories, as the administrative right is, the present work means to deal with these new technologies of which is expected to grow the use in the public administration that will pass inevitably from the bureaucracy to the office automation. The present work is structured in the following four chapters. The first chapter deals with offering a synthetic painting about evolution of artificial intelligence, with a space dedicated at the classification and qualification of the emerging technologies in their interconnections with the public administration, in the usage to speed up and simplify its processes and its activities, also thinking about the imputability of the automated decision and correct application of the computer science rules applied to the algorithm. The second chapter, means to analyze the juridical principles whose respect is requested in the use of tools endowed with artificial intelligence. And, after reviewing the fundamental principles of the administrative procedure, and related at the algorithmic decision, you may not consider the international context in which our legal system is inserted. The third chapter intends to observe the adjustment attempts on behalf of the European Union of the phenomenon, analysing in the detail the Parliament 's Proposal for a European Regulation and Council that establishes rules harmonized on artificial intelligence and modifies some legislative deeds of the Union. Finally, the fourth chapter is dedicated to the future prospects. It will try to examine the spaces, the limits and the risks that arise with the introduction of the algorithmic instrument in the administrative system. In particular it will try to evaluate a concrete usage of the AI systems in administrative method as possible risk of a minor boundary line between machine and person in the figure of a robot as the mouth of the law. The digitization process that is passing through the actual society has a constant impact of growth about the administration and the way in which it operates. This process not only involves the dematerialization of the public activity and relationships between administrative and private authorities, with the aim of simplification and modernization, but brings with it also the use of tools and digital technologies that modify the power and the ways of its exercise. The equipped systems with AI, however, beyond that provide to the public authorities significant gains in terms of efficiency and knowledge, thanks to powerful analytics tools of the data and prediction, they carry with them some risks, on which the attention of the Italian doctrine and beyond has been focused for a long time. Currently the risks are multiple for privacy, security risks of the citizens, for the fundamental rights and for some democrats principles. Also, there is no certainty that the systems equipped with AI are compatible with the fundamental principles of the action administrative, such as transparency, motivation, participation and impartiality due to their opacity and difficulty that they present in understanding them. It seems legitimate to wonder if a nucleus of institutes and fundamental principles built around the human official may be applicable to a PA called to decide through algorithms and, therefore, through the use of “intelligent” machines. Currently different degrees of advancement are recorded in the use of AI systems with reference to central administrations such as INPS and Agenzia dell’entrate, in fact, these institutions are united by a very advanced digitization process and by experimentation with established chatbots and AI systems, both in anti-fraud operation, in the fight against fiscal tax evasion. The ministry of Justice, on the other hand, is characterized by a slow and uneven digitization process in the different sectors of the justice, with evident relapses on possible use of systems equipped with AI. However, one of the causes of the delay in experiments about the use of new technologies for the administration of justice is its decentralization that allows it to operate only on limited samples of data. The administrations are experimenting AI tools outside a normative framework that clearly defines their use within the procedure and the role of the decision human maker. Besides, it seems to be absent an internal policy aimed at equipping administration to the human surveillance, starting from the training staff, which needs to be able to understand logics and the operation modes of the AI systems, thus promoting engagement of the private individuals, whose management skills of the systems result to be strengthen. The latter question is intended to acquire a major relevance in the next future, as proposed in the regulation of the European Commission, in line with what already established by the GDPR, identifies in the “human surveillance” one of requirements required for the use of the AI systems. On the relationship between the tools equipped with AI and the administrative process, according to authoritative doctrine there is the possibility of a complete replacing of the judge with a machine, though they are in favor for the usage of it as an assistant to the human decision-maker. The judge needs to keep full autonomy of judgment and must not base his decision exclusively on directions coming from a computer. Therefore, there is an orientation that pushes for the usage of the machine as an aid to the human judge, assigning it the role of a sort of general robotic lawyer, that will have the task of providing the human judge technical evaluations, the reconstruction of the facts, as well as preparing drafts of the measures and for the treatment of serial, repetitive cases. So the human judge, in total autonomy, will assess this decision or disregard it but giving reasons (Ruffolo, 2021). Therefore, a system is proposed that preserve the primacy of the human rights decision and the absolute freedom of judgment of the human judge, placing the machine side by side with a servant function, which will provide a draft sentence that the human judge will be free to disregard, but decisively justified. For some years administrative jurisprudence has faced this theme by means of the application of the general principles and in particular through the so called algorithmic reasonableness. Some judgments of the TAR have been significant in which the administrative judge has recognized lack of preliminary investigation, lack of motivation and misrepresentation of the facts as at the basis of the analysis there were textual correlations between inconsistent data with the scientific assessment, which combined with other indexes have led to distort the final result. Also, from hermeneutics adopted by some rulings of the State Council, emerges as the administrative automated decision requires the judge to assess the fairness of the IT process in all its components: from its construction, the insertion of the data, to their validity, to their management. Indeed, in the event of a decision founded on an algorithm, a declination is ensured strengthened by the principle of transparency, understood as full knowability of the rule expressed in a different language from that juridical one. Furthermore, since it is a decision made by a public administration, also highlights the art. 41 of the Charter of Nice, and the right to a “good administration”, where it is claimed that, when public administration means to make a decision that can have adverse effects on a person, it must hear him/her before acting, to allow him/her the access to its archives and documents, as well as to “give its reasons for its decisions”. The judge’s decision must be justified moreover, so that the parties understand the reasons but also to allow to be contested before a superior judge. Any decision implies a choice and any choice presupposes a selection of values and purposes. If this applies to human being, a doubt arises for the eventual algorithm that it seems decide without choosing, but applying a series of calculations for the solution of the problems. However, neither the automatic decision really is neutral as to operate before and for the operation of the machine there is always the man, who acts on the basis of options of value and/or interest. This aspect is of is crucial importance for the analysis of the theme as the charm of the algorithm resides on the belief that the perfect measure is that free from errors and that, not to commit mistakes, no may be human; that the car is not influenced, upset, polluted by emotions or matters of conscience and that it is not susceptible to corruption; that it makes sure efficient results, since it reduces response times and so on prevents slopes; that offers effective results because more suitable to the care and to the satisfaction of the pursued goals, guaranteeing impartial solutions. But the decision qualitatively ideal is not the “perfect” one, but rather the “right” one, where the predetermined rules must be respected and adapt to the subjective situation of the care. Furthermore, efficiency is not a prerogative of the machine, also the man can be efficient when his activity is organized and with suitable resources. Finally, the impartiality of the machine if apparently implemented by the adopted decision, may not be, during the identification of the processing criteria of the algorithm. Those which emerge from the survey, in fact, are administrations that want to experiment innovative technologies to improve quality and effectiveness of their own activity. The administrations, in other words, try to use the AI systems because these allow to innovate radically the ways of carrying out the activity of which they are responsible and the expected results, regardless functional specificity of each administration. Therefore, it is necessary an effective legislative solution, such as a code of technological conduct, in whose context the whole matter is disciplined and in which could merge both technical themes and more juridical issues connected to the technologies that there are in the public administration, to administrative organization that believes to be preferable to the administrative procedure, to the imputation and liability model. Then we must consider that currently, there are no AI systems capable of replacing a lawyer completely. Systems which, however, are able to arrange – with the slightest cooperation of a human being – legal simple or repetitive acts, extract norms, precedent judgments, maxims, arguments, or to analyse documents concerning states of reality and understand them at the equal to an human operator for the limited purposes to which their knowledge is useful. If it is certain that in the next future we will assist at the development of AI systems more and more autonomous and able to take decisions complex in several fields of applications; it turns out essential to understand if ethics problems related to such systems must counter this technological development, or if there is a way to implement a useful and responsible technological development that may bring benefits for humans. In conclusion, a very important role in the AI systems is played by transparency, in fact, the same needs to be secured by the right of access to the algorithm, from its “openness ” understood as the possibility of making manifest and understandable to the man the way to operate of the autonomous systems, the reasons that they have brought to certain choices, and the concerning risks, with the aim of maintaining control of man on the machine , communicate and interact with AI systems and have knowledge of their actions. In view of what has been analyzed in this work it is possibile to try to draw a line of continuity with those that will be future researches on this theme, it will be interesting to understand what the future will be of the central figures in the administrative process (for example the person in charge of the procedure). At the same time, it will be interesting to see the effects of the application of a regulation on artificial intelligence for purposes of the European data governance and control on market for these smart systems.File | Dimensione | Formato | |
---|---|---|---|
TESI_DOTTORATO_INTELL_ARTIF_ALES_PUZZ.pdf
accesso aperto
Dimensione
1.21 MB
Formato
Adobe PDF
|
1.21 MB | Adobe PDF | Visualizza/Apri |
I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/20.500.14242/183981
URN:NBN:IT:UNISTRADA-183981