Respuestas de foro creadas
-
Respuestas
-
Last year, the Office of the Seimas invited external lecturers to conduct several one-off seminars/workshops on AI. This month, an informal Microsoft Teams group “AI enthusiasts” was created to help build the governance structure for AI, think of ways to include AI technologies in the processes of the departments/units of the Office of the Seimas, and share knowledge about matters related to AI, its use cases, and specific tools. Personally, I believe that perhaps the most useful support or training we could offer is to create a detailed document (“bible”) that can be consulted each time a generative AI tool is used. Such a document would contain clear reminders, prompt engineering checklists, examples, and other structured information one could follow to get the best possible result in a particular scenario. Unlike lectures or workshops, it would “hold one’s hand while walking through the AI jungle”, thus reducing the time costs, overwhelm, and hesitancy of working with generative AI and enabling people to learn through practice. Learning by actually trying to fulfill your tasks is better than participating in seminars/workshops since using generative AI is more of a craft that requires practicing as much as you can by yourself. (I have created such a guide (26 pages) for the Verkhovna Rada of Ukraine, but I have not translated it, and it is currently not used by the Office of the Seimas.)
The governance structure is currently being drawn up by the Strategy and Innovation Unit with the help of other departments and units.
As for the internal structures governing the use of (generative) AI, the Research Unit is the only entity in the Office of the Seimas that provides any directions on this technology. In December 2023, our Research Unit approved new internal quality assurance guidelines, which have a paragraph about the use of AI for our work. The paragraph, when translated from Lithuanian, sounds like this:
“The Unit may use generative artificial intelligence (AI) technologies to gather relevant information necessary for the preparation of high-quality analytical work, and the author(s) shall keep the Head of Unit informed thereof. If answering the inquiry and the resulting written research document involves the use of AI technologies, this shall be indicated in the document’s footnotes and references to sources. The purpose of the use of AI tools shall be explained, together with a brief description of the information gathering and content generation process. The department’s work shall ensure the ethical use of the output generated by the AI tools.”As a parliamentary researcher/analyst, I find tools like Perplexity Pro or NotebookLM to greatly aid me in my research. The former tool is especially important for quickly collecting information from foreign sources in languages I do not understand while the latter enables me to find information among tens of carefully selected sources.
Despite not being directly in charge of overseeing AI planning in my parliament, I suggested these applications of AI to the relevant colleagues from the Office of the Seimas:
1) creating a chatbot that would use internal databases of our parliament and the office of the parliament, and perhaps even the database of all former and current Lithuanian laws to answer queries regarding current legislative processes, rules of operation, and any other questions an MP or a civil servant may have ;
2) creating a separate chatbot to answer less-sensitive queries regular citizens have;
3) regarding the safety and privacy of AI usage, writing up guidelines for safe AI use*, as currently only the Research Unit has those (for itself);
4) regarding the efficiency and opportunities of using AI, the Office of the Seimas needs to either offer paid subscriptions to at least one or two AI tools (the free versions are limited either by restricted functionality or usage limits) or, preferably, dedicate a budget that could be used to pay for AI-related needs that arise;
5) developing an in-house AI system that would simplify law-making processes by finding similarities with existing legislation, identifying possible conflicts with the law or court decisions, and even offering insights into the possible impacts;
6) thinking of how to achieve new and old ambitions of the Office of the Seimas in the light of the new era of no-code solutions that create opportunities for developing applications cheaper, faster, and without the need for advanced skills.* Guidelines created by Dr. Fitsilis and two other co-authors can be a blueprint for a shortened local version: https://www.wfd.org/sites/default/files/2024-07/wfd-ai-guidelines-for-parliaments-2024-english.pdf
In a broad sense, AI in parliaments should be viewed as an eliminator of human errors and a vehicle for automatization. It could (and should) enable offices of parliaments and even elected members of parliaments to find inconsistencies and deficiencies in their work, as well as spot conflicts between new initiatives, proposals, or documentation and existing law, documents, or rules of operation without the need of painstaking human cognitive processes. Also, much of the work in the offices of parliaments could at least be speeded up by partly automating portions of tasks various departments face, especially the tedious and/or low-value parts of those tasks. More optimistically, AI (agents) could double or even partly replace some costly deliberations by simulating democratic or administrative debates on even the most minor topics, for which real public consultations would not be worth the cost or effort.