Összefoglalás.
A tanulmány a terrorveszély felismerésének és kezelésének elméleti és technikai
összefoglalását nyújtja. Kiemelten foglalkozik a kockázatbecslés technikájával,
és bemutat egy új, mesterséges intelligencián alapuló eljárást, amelynek
segítségével 90%-os sikereséllyel lehet azonosítani azokat, akik
terrortámadásokat hajthatnak végre. E mellett, az eljárás által szolgáltatott
adatok felhasználásával sikerült leírni a radikálisok két típusát, amelyeket
megközelítőleg az „alárendelt/erőszakos” és a „vezető/nem-erőszakos” szavakkal
jellemezhetnénk. A két csoport között jelentős különbségek voltak a családi
háttér, iskolázottság, radikalizálódási folyamat, állampolgársági múlt, bűnözői
előélet, és szerepvállalási jellegzetességek tekintetében.
Summary.
Although Hungary is in a privileged position regarding the threat of terrorism,
the history of other countries suggests that similar good positions can be
temporary. The threat of terrorism can be investigated by several scientific
approaches. After reviewing these, we analyze the theoretical and technical
background of risk assessment, and present the results of our recently concluded
research. In this we examined the US database PIRUS, which contained 112 types
of personal data of 2,148 radicals. About half of them did carry out terrorist
attacks the other half did not. Based on the individual characteristics of the
radicals, the XGBoost machine learning algorithm correctly identified the
perpetrators of the terrorist attacks with a probability of 87%. By using the
data provided by the software, it was also possible to describe two types of
radicals, which could be roughly characterized by the words
“subordinate/violent” and “leader/non-violent”. The former usually had a
criminal but not a radical background. They converted late in life (if their
radicalization was of a religious nature) and adopted radical ideas as adults
(if their radicalization was nonreligious in nature). They played a subordinate
role in terrorist groups, required training and were largely influenced by
social media. They also belonged to low social classes and had many personal
problems. In contrast, non-violent extremists were characterized by a family
tradition of radicalism, mostly had no criminal past, belonged to higher social
strata, and played leading roles in terrorist organizations. Instead of
committing attacks, they engaged in illegal activity by supporting terrorist
organizations. The two main types probably consist of subtypes. Compared to
violent extremists who were radicalized in prison, for example, those who were
not radicalized in prison were mostly foreigners, were often unemployed despite
their higher education, and compared to those radicalized in prison, they
committed lesser crimes before radicalization. Similar subgroups occurred in
both main groups, but their detailed characterization requires further research.
Our findings suggest that artificial intelligence can become a good tool for the
risk assessment of radicals concerning their proneness to perform terrorist
attacks. Moreover, the risk assessment tool employed here may be useful in
typifying radicals, and studying their radicalization routes.