Research

ADM in Portugal is still at an early stage. A lot of work has already been done, and there is a lot of knowledge in the field. However, in general, citizens do not yet feel the presence of ADM in their daily lives. This is either; because it is not present at all or, when it is present, it is not obvious and/or it plays a limited role.

In recent years, new companies have emerged that mostly aim to work with and sell to international markets. These companies use decent AI tools provided by local universities and build upon scientific knowledge from the field and the human resources sector. Some of these companies, although they are not well known in Portugal, have achieved significant commercial success abroad. This has led to increased attention from the national media who are interested in the activities of these companies and the potential of AI for the country.

Unlike other countries, where the use of ADM has led to public criticism and debate, in Portugal, the absence of ADM processes – or the lack of awareness of them – has so far meant that there has been little public debate on the issue. The media often writes about AI from a critical angle, pointing out the dangers and challenges. However, the coverage mostly focuses on events in other countries which can seem distant to the reality of life for Portuguese people.

This might change soon, as many projects are starting to develop, or are already being implemented, in the country. At the same time, the government has taken the lead by presenting the first National Strategy for Artificial Intelligence and by promoting programs to support AI initiatives in Public Administration. The government is also encouraging Portuguese companies to develop in the field of AI and ADM and to take advantage of opportunities abroad. The prime minister, António Costa, recently traveled to India where he visited a Portuguese company working on a pilot project for the installation of biometric face identification technology at New Delhi International Airport. In addition, Costa has encouraged Portuguese companies to take advantage of the economic opportunities offered by India, especially for engineering and tech companies in this sector, as India is currently building one hundred new airports.

Banks and financial service companies are promoting services that they claim give their clients an immediate decision on their personal loan requests. For example, BPI’s “immediate credit” solution states that clients can simulate the desired loan conditions and, upon submitting their data, “know immediately if the funding is approved”. However, the small print indicates that there is a human decision before the credit is granted.

Younited Credit Portugal also promises an immediate first answer with their online system.

Puzzle, from Banco BNI, advertises a one-minute response to online loan requests. The small print states that it will take 48 hours if the client uses a manual process instead, although it is not clear what that means.

In early 2019, Jornal de Negócios reported that Caixa Geral de Depósitos, the public bank founded in 1876 and one of the biggest banks in the country, expected “automated credit decision (services) to be coming soon”.

In January 2020, for the first time, the Data Protection Authority (DPA) has given an opinion on the use of AI in video surveillance systems. It refused to give its approval in two projects that would implement such systems in public spaces in the cities of Leiria and Portimão, as requested by the police. This system would allow the use of technology “to search for people by physical characteristics, gender, clothing or hair colours”. However, the DPA pointed out the lack of justification for the use of this technology. This conclusion refers not only according to proportionality criteria and taking into account the amount and type of information that can be collected but also to the opacity of the process of pattern analysis. The DPA strongly criticized the projects, as they involved “large-scale systematic monitoring and tracking of people and their habits and behavior, as well as identifying people from data relating to physical characteristics, being undeniable the high risk to the fundamental rights to data protection and respect for privacy”. The DPA also considered that there was a risk of “not being able to understand whether the results presented by the system, which would serve as (the) basis for the police to make decisions about the targeted citizens, are discriminatory”, pointing out that this would be “inadmissible” under the Portuguese Constitution. (Sources: TSF, Público).

In January 2018, Diário de Notícias reported on the use of face recognition by the Laboratory of the Forensic Science Police to monitor foreign citizens in airports. However, it remains unclear the exact context in which such a system is used. The involvement of a forensic laboratory from the police might indicate that such technology is not used as a large-scale standard solution for the control of foreign citizens, but rather as an alternative resource, available to the police, for investigation purposes.

Under the Portuguese gambling law (DL n.º 422/89, de 02 de Dezembro), specific people can be banned from casinos. This includes the possibility of self-exclusion, to help people who are struggling with gambling addiction. Casinos cannot allow persons on the non-authorized list to enter their premises. Traditionally, this law is enforced by the security staff of the casinos using CCTV and personal contact for recognition purposes. In this context, the rise of face recognition was seen as an opportunity to improve the efficiency of law enforcement.

In 2016, Turismo de Portugal (the national tourism board) contracted IBM for 110 licenses and the central software needed to activate and run the face recognition modules of the CCTV systems installed in casinos. The contract cost 337,833.63 euros (plus VAT) and was comprised of 401 CCTV cameras which were set up in all 11 casinos in Portugal. According to the publicly available contract, the implementation period should have taken 90 days.

In August 2018, the Jornal de Notícias newspaper reported (paywall) that casinos were installing face recognition systems intended to stop, for example, self-excluded gambling addicts from entering the premises.

However, in May 2019, the Público newspaper reported (paywall) that the system was still not up and running. The article cites a representative from a casino who confirmed that the system did not produce the expected results, “for technical reasons“, but he thought that such a goal may still be achievable. A representative from another casino confirmed the claim, saying that the system “does not work very well, (but it) is in the implementation phase”. Público further reported that the system was installed back in 2017, but it has not been operational since then. The technical difficulties are related to the poor lighting at the entrances to the casinos, and to the fact that the cameras need to zoom in on the customer’s face. The failure of the system meant that the casinos had to hire more staff to do this work. Tuuismo de Portugal did not reply to questions concerning the renewal of licenses and their plans to continue the contract with IBM. Público also reported, back in 2017, that it was said that the CCTV cameras would only work with IBM software. This would explain why the contract was signed directly with IBM instead of going through a normal public tender process.

Over the last few years, algorithms have played a key role in detecting, investigating and prosecuting cases of fraud associated with medical prescriptions within the national health service (Serviço Nacional de Saúde). Sábado reported how a team of three or four people processed between six to seven million prescriptions a month, written by 40,000 medical doctors and picked up at the 3000 pharmacies throughout Portugal. The team uses algorithms to detect patterns and to flag cases of higher fraud probability. For example, the system checks to see if there is a large concentration of prescriptions from a single doctor filled in at the same pharmacy if a doctor prescribes above-average quantities, which doctors give out the most prescriptions, or who gives the most value, etc.

This information is then forwarded to the criminal investigation police. Between 2012 and 2016, the reports forwarded for police investigation involved 432 doctors, and more than 122 service providers and prescriptions worth a total of 1,098 million euros. The police then investigated which of these cases might have illicit origins, which resulted in several police operations and subsequent criminal prosecutions with significant impact in the press and with the public.

Every year, approximately 11,000 retired Portuguese citizens who live abroad must demonstrate “proof of life” in order to continue receiving their pension. Now, a pilot project launched by Caixa Geral de Aposentações, a welfare institution for civil servants, aims to simplify the process by allowing pensioners who live abroad to demonstrate the required proof of life via face or voice recognition. The pilot project started in Macau (a former colony that remained under Portuguese administration until 1999), where hundreds of people are already using the system.

A representative of a pensioner’s association in Macau told Antena 1 that, while he found the system easy to use, he also heard criticism from other pensioners who had trouble with it.  The problems were mainly related to the fact that these are senior citizens who are not familiar with computers and, in some cases, even mobile phones.

In December 2019, Uber announced that it had started using its new driver identity verification and validation system in Portugal. The company said that “the new feature will require drivers and delivery partners to take a picture in ‘selfie’ mode before going online. The picture is then immediately cross-checked with the profile photo of the driver or delivery partner and so we make sure you are the right person behind the wheel or making the delivery”, Expresso reported. The Real-Time Identity Verification system was already being tested in the United Kingdom, after the passenger transport platform faced problems in London, with the local taxi regulator refusing to extend its license due to safety concerns.

In 2019, the Portuguese Immigration and Border Service spent 4.2M euros + VAT to update Project RAPID (Automatic Identification of Passengers Holding Traveling Documents), a system that started in 2007 and that automatically controls passengers who cross Schengen’s external borders. “This system combines the operations of reading and checking electronic passports with an innovating feature for assessing biometric data which operates an automatic gate opening device. This device checks on a first phase the genuineness of electronic passports and validates all data stored in the chip and, on a second phase, appraises the passenger’s identification by establishing a comparison between the photo stored in the chip and the information of the passenger in loco, automatically opening the border gate when the features of both images are coincident”. The recently purchased devices will be installed in the airports of Lisbon, Porto, Faro and Madeira (source).

During October 2019, the Humberto Delgado International Airport in Lisbon tested a voluntary trial called Biometrics on the Move. This is a Frontex (the European Border and Coast Guard Agency) project together with the Portuguese Immigration and Border Service and the Lisbon Airport Authority (ANA). It uses face recognition and touchless scanning of fingerprints enabling passengers to cross the border “almost seamlessly”, “without even taking out their passport or other documents”, according to Frontex.

Vinci Airports, a company that develops and operates airport platforms, chose the Humberto Delgado International Airport in Lisbon to pilot a center of excellence on flow management, within the network of airports the company operates.

The center of excellence in Lisbon is currently developing innovative projects related to a barrier-free system, which will allow a seamless flow of passengers from the entrance of the terminal until boarding the flight, without having to stop or show documentation. In order to use this system, the passenger will have to submit passport data and allow a photograph to be taken. Then, biometric reading devices (fingerprint or face recognition) scattered throughout all the necessary stages will read this data as the passenger is moving, and allow him or her to pass without stopping.

The company intends to start implementing the new system in Terminal 2, which is used mostly by low-cost airlines, but there’s no date set for the implementation.

“Operação Marquês” is the nickname of a highly controversial judicial process where the ex-prime minister of Portugal, José Sócrates, was detained (preventive detention) while facing charges of corruption, qualified tax fraud, document forgery, and money laundering.

Following an indictment at the investigation phase, the defense requested the opening of an optional phase (“Instrução”). This phase takes place before the trial phase and allows the investigating judge to assess the consistency of the evidence gathered by the public prosecutor. The judge can then determine whether the evidence is sufficiently robust to lead to the possible conviction of the defendant. For this specific case, this optional phase would fall under the jurisdiction of a specialized court that has only two judges, one being the judge that oversaw the investigation phase and made decisions that the ex-PM and his lawyers heavily contested in public. They also argued why the same judge who oversaw the investigation should end up being the one to lead the next phase of the process.

The selection of the judge in charge of the next phase became a contentious topic. Public opinion on which judge should be selected was divided and often influenced by personal views on the ex-PM. Some argued that the same judge who was thought to be tougher should continue. Whereas, others preferred to have the alternative judge, who, in their view, was fairer. In the end, it was left up to “the algorithm” to randomly “choose” which judge it would be.

A TV cameraman and a photojournalist were brought into the room as the random algorithm decided who would be the judge to take the case. This led to public criticism from the first judge, effectively putting the algorithm under suspicion. According to him, the algorithm was not purely random as it took into account several factors, such as the number of opening cases each judge had been involved in, and, as such, it could be manipulated, to a degree. This led to the opening of a disciplinary process for the judge, later filed. In a press release, the independent judicial entity responsible for the management and disciplinary action of judges said that “the electronic distribution of cases is always random, not balancing on a daily basis or in any other time period, that may be known in advance, the cases distributed to each judge”. However, a fact-checking agency has confirmed from several sources that the system is not entirely neutral and that it is influenced by external factors, including the number of cases assigned to each judge.

None of the criticism has ever focused on the fact that true randomness is actually very difficult to generate in a computer system or the fact that the algorithm is not public. In fact, very little is known about the algorithm, as the previous contradictions show. For example, it is not known if it was ever audited. The judicial process is still ongoing, and it is still not known if the ex-prime minister, José Sócrates, will face a trial.

In 2017, Expresso reported how human resource companies in Portugal were taking the first steps in automation. At the time, Manpower confirmed that they were not using it in the country yet, but that they were preparing for such a possibility. Randstad confirmed that they were using assessment and applicant’s validation tools. Expresso says that despite the clear and fast disruption of the human resources sector in the EUA, change has been slow-paced in Portugal, “both because of legal issues (such as data protection) and because the human component continues to be highly valued in this market”.

With one million euros in funding from the European Union (under the Connecting Europe Facility), the project Urban Co-creation Data Lab intends to respond to five challenges in the area of municipal management: micro-mobility, waste, parking, pollution, and crowd management. The project started in Lisbon, but it will also be tested in two other European cities which have yet to be selected. According to the project’s leader, the main objective is “bringing together the best European brains in urban analytics to collaboratively create models of artificial intelligence to support decision-making in municipal planning and management”. (source)

In 2019, at the Web Summit, the Institute of Records and Notaries (IRN) presented Irene – a virtual assistant whose mission is to help users on matters regarding the Citizen’s Card. At the event, IRN’s vice president described the capabilities of the service: “In addition to answering a set of questions, Irene engages in a conversation in which she asks a set of questions to the citizen, in order to understand their particular situation and thus be able to create an optimal experience in the renewal of the Citizen’s Card”. The system is currently being tested and is available on IRN’s website.

Founded in 2017, Heptasense offers a video analytics service for surveillance cameras to automate and optimize security management. The company claims to be “against facial recognition” and believes that “video surveillance systems have always taken this path of espionage, of invasion of privacy”. Their idea is to detect suspicious behavior without face recognition turning conventional surveillance cameras into intelligent systems. In their own words: “the system is able to analyze a person’s behavior and thus anticipate its possibly illegal or suspicious actions, helping the security to prevent dangerous incidents”, adding that “it is the first AI software that learns and understands human behavior patterns, mimicking the way the human brain works and identifying potential threats to safety and security”.

It is not clear if this is the system used in the previously mentioned projects in Leiria and Portimão related to the use of AI in video surveillance systems, although there are similarities in the description of the features. However, the criticism put forward by the Data Protection Authority also applies, namely the issues concerning fundamental human rights and the opacity of the system. (please see “Police Forces” section).

In 2019, the government presented the National Strategy for Artificial Intelligence. The document was created by the Ministry of Science, Technology and Higher Education and its main objective is to make Portugal part of the European network of centers of excellence in artificial intelligence, in the short to medium term. 

The Action Plan includes seven points of action:

1. Inclusion and Education – widespread dissemination of knowledge in AI,
2. qualification, and specialization,
3. thematic areas for research and innovation in European and international networks,
4. public administration and its modernization,
5. specific areas of specialization in Portugal with international impact,
6. new developments and support areas in European and international networks, and
7. new challenges of society brought by AI: Ethics and Safety.

In 2018, the government launched a program to support new R&D projects in a partnership between the Public Administration and scientific institutions: “The projects should be oriented to deepen the processing of public data and stimulate the production of new knowledge relevant to citizens, based on the use of advanced techniques of artificial intelligence and data science”.

In 2018, the following projects were the winners: Use of Artificial Intelligence to Enhance Teledermatological Screening; Data Science for Emergency Medical Service Optimization; Identification and Prediction of Hospital Emergency Demand; Ground Recognition System. A new edition of the program was launched in 2019.

While there is no specific law or oversight entity for ADM processes, ADM might be relevant for the activity of sector regulators.

In July 2019, the Competition Authority alerted companies that sell goods or services online to the possible implications of the algorithms they use. In its report, the authority included a survey of 38 companies with an online presence in Portugal. 37% of the companies surveyed admitted using algorithms to monitor the prices of competitors. Of those, 78.6% said they adjusted prices according to the results, and 7.9% of those made those adjustments automatically, and without human intervention.

The regulator considers that “the monitoring of competitors’ strategies and the adjustment of the price or other strategic variables based on the information collected have always been part of the normal functioning of the markets”. However “their frequency and extent intensify in the digital economy”, and so “the frequency of their use can facilitate collusion strategies”. It adds that ultimately companies “are responsible for the algorithms they use”.

Founded in 1984, the main purpose of the Associação Por tuguesa Para a Inteligência Artificial (Portuguese Association for Artificial Intelligence, APPIA) is “promoting research, teaching and dissemination of artificial intelligence as a scientific field”. With over 300 senior researchers as members, APPIA has long been an active voice in the field, especially within academia. APPIA organizes multiple events around AI, the main ones being a yearly AI conference and an advanced course on AI.

Calouste Gulbenkian Foundation is a Portuguese institution dedicated to the promotion of the arts, philanthropy, science, and education. As a major reference in the promotion of such areas, especially through funding, the foundation plays a vital role in the areas of activity it covers. Furthermore, Gulbenkian offers scholarships intended to attract students from Portuguese higher education institutions to research in Artificial Intelligence.

D3 is a digital rights organization, and a member of the European Digital Rights network. The organization has voiced its public position against the use of artificial intelligence systems by police forces for public surveillance purposes.

Portugal seems to be in a very good position to develop its ADM capabilities, despite being far from a world leader in the industry. As a result of extensive academic work, together with a strong, well prepared technological sector and competitive wages, there is the potential for unique growth opportunities for the small and medium-sized companies that have come to dominate the Portuguese business sector in recent years. This is increasingly felt by various actors, especially those in both the business and political arenas. The creation of the National Strategy for Artificial Intelligence by the government is one such example. However, there is a feeling that, while the economic opportunities of AI are immense, there is a limited time-frame, which cannot be missed. The main economic barriers are the usual difficulties of the Portuguese economic fabric relating to access to capital and the lack of scale of most enterprises.

On the other hand, public criticism and debate related to the dangers and opportunities created by ADM processes remain limited. The fact that in Portugal, many of the controversial solutions adopted in other European countries are not present has meant that citizens have not yet felt the need to bring these issues up in public discussion, despite the fact that both the public and the press follow developments abroad very closely. Regardless of this fact, the research shows that new initiatives are emerging very rapidly in Portugal, within the AI and ADM sectors. However, most of these initiatives are still in the very early stages. Therefore, even when ADM mechanisms are in place, citizens may not feel their presence as they might have a very limited application for small specific groups within the larger population.