• No results found

Most, if not all surveillance technologies have an impact on privacy and liberty to a certain degree, some by themselves, some in combination with others. This can also be said of agent technology. Based on the conclusions of the previous chapters, particularly chapter 4, I believe that agent-enabled sur- veillance will have a significant impact on privacy and liberty in the (near) future.

Moreover, I believe that the impact of agent-enabled surveillance on privacy and liberty will be mainly of a quantitative nature. That is to say, the use of agent technology for surveillance will have a measurable effect on privacy and liberty, but the application of agent technology as such does not form a break with past surveillance practices and the possible threats they pose. Arguably, this means that up to a certain level the existing legal framework is adequately suited to deal with the quantitative effects of agent technology. Below I shall describe five quantitative effects that agent technology will have on surveillance and thus on privacy and individual liberty. They are: (1) more efficient data monitoring and data gathering (subsection 7.1.1), (2) more effective data exchange and data mining (subsection 7.1.2), system integration (subsection 7.1.3), empowering surveillance operators (subsection 7.1.4), and replacing surveillance operators (subsection 7.1.5). These five quantitative effects will be handled as follows. In chapter 7 they will be dis- cussed, in chapter 8 they will be reviewed according to the current legal framework, and in chapter 9 recommendations will be given for the enhance- ment of the legal framework.

7.1.1 More efficient data monitoring and data gathering

A first quantitative effect of agent technology is more efficient data monitoring and data gathering. We have seen that information overload is one of the

primary reasons to use agent technology. By using software agents, surveillance operators and law enforcement officers can overcome the information overload. Tools that help overcome the information overload are particularly useful on the internet. The internet houses an enormous amount of information. Unfortunately most of this information is in unstructured form (i.e., natural language). This means that it is difficult to extract useful information from all the available data in an efficient matter. By using agent technology law enforcement agencies can search larger parts of the internet for information that is relevant to an investigation. In addition, agents that are able to interpret natural language can make better sense of unstructured data on the internet. When the semantic web becomes a reality, software agents will become even more effective at monitoring and gathering data.

Furthermore, since software agents are particularly well suited for unobtrus- ive monitoring, they can be used to monitor data sources continuously online. Obviously, this development will make surveillance and monitoring far more effective, and threats to privacy and liberty more likely.

7.1.2 More effective data exchange and data mining

A second quantitative effect of agent technology is its contribution to more effective data mining. In particular in the area of (distributed) data mining, agent technology can have a profound impact on the efficiency with which data is integrated. As we have seen in chapter 4 software agents can be used to integrate heterogeneous, disparate, and geographically distributed databases into a single ‘virtual database’, making data mining (and thus surveillance and control) far more effective.

We have established that technical barriers to accessing and processing personal data lead to inefficiencies that act asde factosafeguards for privacy and liberty. The protection of privacy and liberty in the information society is for a large part dependent on these inefficiencies. Where there are no techni- cal barriers to accessing and processing personal data, the legal framework puts in place artificial inefficiencies in the form of regulation that forbids aggregation and integration of databases. So, on the one hand there is a need to connect the dots, while on the other hands there is the notion of a free society that seeks to keep the power to connect the dots away from any one actor, particularly the central government (Taipale 2003, p. 58).

We have established in chapter 3 that we can distinguish between two approaches in data mining, viz. subject-based inquiries and pattern-based inquiries. Both influence privacy in different ways, so I shall describe their effects separately.

Subject-based inquiries and privacy

Agent technology enables surveillance operators to query many databases simultaneously in search of information regarding a specified individual. Furthermore, software agents can also search the internet for information regarding a specified data subject (the idea of dataveillance). So, we may conclude that through agent technology subject-based inquiries will become more efficient, effective, and complete.

Extensive aggregation and integration of data can ultimately lead to what Taipale (2003, p. 58) calls ‘the demise of practical obscurity’, a situation where an individual is unable to hide himself from outside scrutiny. Such a situation is detrimental to an individual’s privacy and liberty. As described throughout this thesis substantial digital dossiers regarding individuals can be used to exercise (social) control. The more efficient the aggregation and integration of personal data becomes (for instance through agent technology), the more profound its effects on privacy and liberty will be.

Apart from the exercise of direct control through knowledge gained from a subject-based inquiry, there are also panoptic effects associated with data aggregation and integration. Extensive aggregation and integration of personal data can have a chilling effect on individual behaviour (Minowet al.2004, p. 35). As a result of extensive government oversight or the possibility thereof, individuals might refrain from certain social or commercial activities, mask their behaviour, or reduce their participation in society out of fear of judge- ment. The risk is not only that social or commercial activities are chilled, but also that rights such as the freedom of expression, protest, association, and political participation are affected as well (Minowet al.2004, p. 35). What must be noted is that for panoptic effects to occur it is not actually necessary for agent technology to be as effective as suggested. The mere idea that software agents can gather information is enough to cause panoptic effects. In other words, knowledge is power, but potential knowledge can equally present power (Minowet al.2004, p. 35).

Pattern-based inquiries and privacy

Software agents can also facilitate pattern-based inquiries through distributed data mining. The problem with pattern-based inquiries from the perspective of privacy and liberty is that the data analysis is not based on an individualised suspicion. In other words, everyone is a potential suspect. If your behaviour matches certain criteria, you automatically become a suspect. This kind of pro- active searching for criminal behaviour that has yet to take place is reminiscent of Dick’s novelThe Minority Report, in which mutated humans called pre-cogs are able to predict future crimes (Dick 1956, p. 71-101).1In the book the pre-

1 In 2002The Minority Reportbecame a major motion picture starring Tom Cruise. Though the plot of the movie differs somewhat from the book, the central theme of surveillance and control remains.

cogs are able to witness a murder before it actually takes place, thus making it possible to apprehend the potential killer in advance. The obvious effect this has is that premeditated killings become a thing of the past. The potential panoptic power that flows from this ‘precognition’ is evident. While these effects might be desirable in the case of murder or terrorism, there are many situations where such power is less likely to be desirable. Of course, pattern- based inquiries do not have the same level of precision and accuracy as the technology described inThe Minority Report, but the idea is more or less the same.

The idea that the government is able to match certain behaviour auto- matically to a ‘suspicious pattern’ can lead an individual to adopt a be- havioural pattern that is more consistent with the perceived social norm. By defining which behavioural patterns are considered deviant, the government can effectively regulate individual behaviour. For instance, it is possible that people will alter their behaviour to avoid being considered a potential suspect by refraining from conducting activities that might be categorised as ‘high risk’ activities. So, when pattern-based inquiries are used for the purpose of regulating behaviour positive liberty is at stake. The use of pattern-based inquiries might be justifiable for the prevention of terrorism and organised crime if adequate safeguards are in place, but there is always the risk of function creep.2Moreover, as the behavioural patterns of political and social

minorities are more likely to be labelled as suspicious or deviant, there is also the risk of discrimination and social inequality.

7.1.3 System integration

A third quantitative effect of agent technology is its contribution to system integration. In section 3.2 I have described the rhizomatic expansion of sur- veillance throughout society. Our growing reliance on surveillance for security, efficiency, and convenience has prompted the installation of many different surveillance systems and infrastructures, both in the public and in the private sector. Though surveillance systems of both a liberal and a disciplinary nature are being deployed by a variety of actors without a centralised, hierarchical structure, there is a strong desire (especially on the part of the government) to bring these systems together and integrate them into a larger whole. In general, surveillance systems are built with a specific purpose and surveillance

2 An example of function creep can be seen in the MATRIX (Multi State Anti-Terrorism Information eXchange) programme. The MATRIX was a tool for large-scale data-exchange commissioned by several states in the United States in response to the September 11 terrorist attacks. It combined information from a multitude of public sector and private-sector databases to aid law enforcement officers in the identification of terrorists. However, according to figures issued by the programme itself, just 2.6 per cent of all the queries to the system were actually related to terrorism (Seifert 2006).

domain in mind. Typically, these systems employ only a single type of sur- veillance technology (for instance, CCTV, RFID, orGPS). Consequently, sur- veillance systems are not developed with interoperability in mind. However, combining different kinds of surveillance mechanisms and integrating different surveillance systems into a larger whole can greatly enhance the overall effectiveness of surveillance, in particular in the area of (automated) monitor- ing. Information and communication technology plays an important part in this development, as it allows for the rapid expansion and integration of surveillance (Lyon 2003b, p. 94).

Among the technologies for integrating surveillance mechanisms and systems, software agents feature prominently. As I concluded in section 4.5, software agents provide mediation services and query brokering to surveillance operators by enabling logical and semantic interoperability of (previously) discrete surveillance mechanisms and systems. The most striking examples of integrated surveillance systems are the sensor networks and (collaborative) decision-support systems described in chapter 4. These sophisticated sur- veillance networks provide a glimpse into a possible future for surveillance and control. Furthermore, they illustrate how powerful ‘next generation’ surveillance might be and how it will influence privacy and (individual) liberty.

The lack of interoperability between different surveillance mechanisms and systems acts as ade factosafeguard against excessive surveillance powers. However, as we can judge from the development of distributed data-mining systems, collaborative decision-support systems, andC4ISRsystems, ongoing system integration made possible by software agents is rapidly removing technological barriers to system integration. By combining different types of surveillance systems into a larger whole a more complete surveillance infra- structure can be created. Naturally, a more comprehensive surveillance infra- structure has adverse effects on privacy and liberty. In a sense, agent techno- logy ‘multiplies’ the effectiveness of (previously) discrete surveillance systems. In this way, system integration will enhance the possibilities for the exercise of control, threatening privacy and liberty. Moreover, system integration will contribute to the panoptic feelings felt by those being watched.

7.1.4 Empowering surveillance operators

A fourth quantitative effect of agent technology is the way in which it can empower surveillance operators. The improvements in logic and semantic interoperability that agent technology provides, makes surveillance operators more effective in general. But apart from improving logic and semantic inter- operability, software agents can also empower surveillance operators by making their job easier. Software agents can make surveillance operators more effective by acting as ‘personal assistants’ (i.e., theCOPLINKActive Agent) to

surveillance operators or by providing decision support (i.e.,COORDINATORS andCOMBINEDsystems). This means that surveillance is enhanced, which in turn may pose an additional threat to privacy and liberty.

The obvious threat to privacy and liberty is the more effective exercise of control as a result of enhanced surveillance. However, it is more difficult to establish if panoptic effects can be associated with the empowerment of sur- veillance operators. Panoptic effects only occur when people are aware of the fact that surveillance is being conducted. I feel it is unlikely that people will realise how software agents will make surveillance operators more effective. If this holds true, panoptic effects will be limited. However, when people do realise how software agents might improve the capabilities of surveillance operators, panoptic effects will become stronger.

7.1.5 Replacing surveillance operators

A fifth quantitative effect of agent technology is that in some instances it can replace surveillance operators. Software agents could thus eventually, at least in part, remove the need for human operators. Examples of monitoring tasks that can be performed by software agents are the autonomous operation of CCTVcameras and surveillance of the internet. In view of the fact that software agents are less expensive in operation than their human counterparts, it is likely that they will be employed more often. This will make surveillance more comprehensive both in scope and duration, particularly in the area of (auto- mated) monitoring. The fact that surveillance will become more comprehensive will no doubt have an adverse effect on privacy and liberty and further strengthen the Panopticon.

As of now it is unclear what the impact of automated surveillance will be on privacy and liberty. At this stage there are not that many applications of agent technology whereby human operators are replaced. The XENON application is one of the first working examples of agent technology taking over the tasks of human operators. In my opinion the impact of automated surveillance will be profound, but in this stage of the development it is too early to draw any definitive conclusions.

7.1.6 Conclusions on quantitative effects

I described how agent technology can contribute to the exercise of surveillance and control. It may be concluded that the quantitative effects of agent techno- logy on surveillance will result in more efficient, more effective, and more comprehensive surveillance. However, this development may also have adverse effects on privacy and liberty when adequate safeguards are not put in place.

The use of agent technology for surveillance purposes is part of a broader development towards the more effective and efficient exercise of surveillance by means of information and communication technology. The quantitative effects of agent technology must not be viewed in isolation, but rather in the light of this broader development. Yet, the quantitative effects of agent techno- logy could accelerate the development of a more complete surveillance infra- structure. Therefore, we may conclude that it is likely that agent technology will further broaden the ‘information divide’ that exists as a result of sur- veillance. I use the term ‘information divide’ to describe the discrepancy in information available to the observers and the observed. Those who have access to advanced agent technology will have better access to information and will more easily gain knowledge, upsetting the balance of power. Agent technology could have a ‘multiplier’ effect on the existing information divide since soft- ware agents are so effective in finding information and making sense thereof for human operators.