Towards a transparent and responsible Artificial Intelligence
By Javier Albusac (University Professor).
We cannot deny it, we are heading towards an increasingly digitized society in which the Artificial Intelligence (AI) it will be an integral part of our social and cultural fabric. Faced with this reality, reflections are urgently needed on how this technology will affect our lives, and if its use will be fair and beneficial to all. If the Industrial Revolution changed the way we produce and consume, and the Digital Revolution transformed how we communicate and access information, the Artificial Intelligence Revolution it will usher in an era where we will work and make decisions in entirely new ways.
AI Explosion and Raising Ethical Questions
This explosion of AI brings the approach of important ethical issues that affect the citizenry. Some of the most common concerns include a) the risk of AI making decisions that harm certain groups of people by accentuating inequalities, or b) the possibility of AI being used maliciously.
In relation to the first concern, it is worth mentioning the case of job recommendation systems for the selection of candidates. Some studies have shown that these systems can perpetuate or aggravate existing inequalities in society if adequate measures are not taken to address these problems. For example, if the system is trained on data that reflects existing gender, race, or sexual orientation inequalities, it may reject qualified candidates just because they belong to certain groups. In these situations it is essential to minimize any potential bias or discrimination, and to ensure that the rights and values of all groups of people are taken into account.
The expansion of AI cannot lead to a loss of rights and guarantees that have taken centuries to achieve with democratic societies.
Regarding the malicious use of AI, it is worth noting its potential for the rapid generation and spread of false news, one of the main concerns in democratic countries. If they are used efficiently, they can generate serious social conflicts and are a weapon for the manipulation of public opinion, which can erode the stability of States and their institutions. Some studies demonstrate the intensification of this type of news during electoral periods, and include causal models to discover how they influenced the results. A second example is the malicious use of surveillance and mass control systems without the proper consent and citizen control. The use of surveillance technologies without a proper legal framework can undermine privacy and individual rights.
Therefore, there arises necesidad essential to design algorithms in which exists transparency and whose decisions conform to the human ethics. In other words, it is important that people can understand and evaluate the reasons behind the decisions made by an artificial system. This is especially important when it comes to decisions that can have a significant impact on people's lives. There are several reasons why it is important that the decisions made by AI are transparent and understandable by humans. First, transparency allows people assess whether the decisions made by the AI are fair and just. If we do not understand the reasons behind a decision, it is very difficult to determine if this decision is fair or not. Second, transparency is essential to ensure the Trust of people in AI. If we don't understand how AI works or why it makes certain decisions, it's very hard to trust it. Finally, transparency is essential to guarantee the responsibility of the AI. If we don't understand how AI works, it is very difficult to determine who or what is responsible for its decisions.
Measures to ensure transparency
Along these lines, there are several measures that can be taken for to guarantee that Artificial Intelligence (AI) be more transparent and that their strategic are understandable by humans:
Six measures for a more transparent and responsible AI- Design socially aware and explainable algorithms: From early design phases, creators should keep this purpose in mind. Design algorithms that are aware of biases and consider the social and ethical impacts of their decisions and actions. These algorithms must be understandable and explainable, including the disclosure of any relevant information that may be necessary to understand how they work and how they make their decisions.
- auditable algorithms: Open access to the algorithms that influence our lives would make it easier for stakeholders to evaluate and verify. This measure can help to minimize any potential bias or discrimination that these systems may have, as it allows experts and other stakeholders to assess and detect potential problems. The citizen must have the right to know the design of the algorithm and its source code, based on the Transparency Law. In this way, each person can also make sure that they are treated by the artificial system as an end in themselves, and not as a means to a particular end.
- Known data sources: Know the origin of the data with which the algorithms are trained and how they are filtered. On many occasions, the latent bias is already found in the data itself, without the programmer himself transferring his prejudices. The data should not belong to companies or governments, but to citizens. In this sense, Spain has been a pioneer through the Digital Bill of Rights, which consists of 25 fundamental rights, such as the right of a person not to be located and profiled, or the right of a person not to be discriminated against by an algorithm.
- Formation of citizenship: As on many other occasions, the solution lies in education. Citizens must, on the one hand, become aware that algorithms are not impartial for the mere fact of being mathematical. On the other hand, you need to be trained to understand how AI algorithms work and how they make decisions. Computing and computer language should already be compulsory subjects in Primary, Secondary and Baccalaureate Education. In some European countries such as Bulgaria, Greece, Hungary or Poland, Computer science is taught as a separate and compulsory subject. In Spain, however, the Ministry of Education and Vocational Training suppressed computer science subjects from the Baccalaureate (formerly ICT). For this reason, some groups started movements to once again reinforce the presence of Information Technology in the curricula of Primary Education, Compulsory Secondary and Baccalaureate.
- Establish rules and regulations: Establishment of rules and regulations to ensure that AI algorithms are used responsibly and ethically. This may include the creation of laws and regulations that establish the responsibilities and obligations of designers and users of AI algorithms. Currently, transparency is an ethical principle recognized by the European Union in numerous documents, such as in the Ethical Guidelines for Trustworthy Artificial Intelligence. At the national level, steps have also been taken in this direction, for example, through the practical guide on the business obligation to provide information on the use of algorithms in the workplace.
- Encourage collaboration and dialogue: It is important to encourage collaboration and dialogue between AI algorithm designers, ethicists and policy experts, and users to ensure that ethical and responsible algorithms are used.
distribution of responsibilities
Who should fall on responsibility for a more transparent and accountable AI?
The answer is: ALL SOCIETY. This includes designers and developers of AI algorithms, who must ensure that these systems are transparent and explainable, and that they minimize any potential bias or discrimination. It also includes companies and organizations that use AI, who must ensure that these systems are used in a responsible and mutually beneficial manner. In addition, it includes regulators and governments, who must establish adequate rules and regulations to ensure the responsible and ethical use of AI. Finally, it includes the general public, who must be informed and educated about how AI algorithms work and how their decisions are made, and who must participate in dialogue and decision-making about the use of AI in society. .
