The processing of information on ‘prison-related living conditions’ is an important instrument for increasing dynamic security in the prison system and is therefore of significant benefit to staff and inmates. This information processing, which is currently regulated in Austria by §§ 15a ff StVG, is to be further developed into an effective prison intelligence system as part of the PALM project.
Currently, relevant questions can only be answered by standard users through time-consuming manual searches of individual screens for several inmates. The ability to run such queries against the existing data in a structured manner is currently reserved for IT experts in development.
This project develops AI-based query functionalities for existing databases, such as in Integrierten Vollzugsverwaltung (IVV) and elektronischen Vollzugsmanagement (eVM). The aim is to simplify and democratise access to and usability of existing systems and improve the use of the information collected to make prison administration more efficient. Such a prison intelligence system serves both to provide more efficient social and medical care for inmates and to identify and prevent security-critical incidents (escapes, smuggling, etc.).
In the future, PALM will make it possible for prison staff to formulate analysis requests in natural language and make the results available in real time in a clearly organised manner. Large language models (LLMs) are ideally suited for this, as they can understand complex language structures and recognise context-related meanings. However, the use of LLMs is limited to the generation of database query statements from natural language queries and the analysis of the data obtained from the generated query, thus minimising potential risks of hallucination and imbalance.
An absolute non-goal of the project is the creation of personality profiles using AI. Similarly, the system will not allow decisions to be made using AI on the basis of inmates' personal characteristics. Rather, the information already recorded manually by prison staff on the basis of legal principles is to be made effectively accessible.
PALM will make a valuable contribution to strengthening dynamic security in prisons in the future by means of prison intelligence, which will benefit both staff and inmates. The PALM project will help to democratise access to the data already available within the prison system and make it accessible to technical laypersons. It is expected that this will lead to a significant improvement in quality and an increase in efficiency in the Austrian prison system.

Project Lead
Dipl.-Ing. Georg Thallinger / JOANNEUM RESEARCH Forschungsgesellschaft mbH

Partners
app informatics zt gmbh
Bundesministerium für Justiz
PKE Holding AG
Research Institute AG & Co KG
Software Competence Center Hagenberg GmbH

Contact
Dipl.-Ing. Georg Thallinger
Steyrergasse 17, 8010 Graz
+43 316 876 1240
georg.thallinger@joanneum.at
https://www.joanneum.at/digital/