I Made Andi Arsana*
Of course, the terms “permissible” (halal) and “prohibited” (haram) in the title of this article are not religious terms to be examined based on the Quran. I borrowed these terms because they were inspired by the name of the Arsana Space event organized by Arsana Ecosystem on February 9, 2025. In this event, Robertus Setiawan Aji Nugroho (Aji) and I were invited to share insights about the use of AI in research. Since it was held in the evening before breaking the fast, the organizers used the terms “permissible” (halal) and “prohibited” (haram).
Aji is an outstanding expert. Not only is he a practitioner of AI, but he has also been conducting research in the field for over a decade. Long before ChatGPT became widely popular, Aji was already deeply engaged with the power of computers to “understand” relationships and “meanings” between human words and sentences. Aji earned his Ph.D. from Macquarie University in Sydney, after previously obtaining his Master’s degree from UNSW in the same city in Australia. When it comes to AI, there is no reason to doubt Aji’s expertise.
I was delighted to be asked to join him in sharing insights about AI with around 60 enthusiastic participants, most of whom were fasting. The event was held online, allowing participants to listen while lying down. Some might have even been ironing their clothes. Who knows?
Dina, the moderator from Arsana Ecosystem, began with introductions and then asked me to share my experience using AI. I recounted an experience when I was preparing to teach a Database Systems (DBS) practicum at the Geodesy Engineering Department at UGM. Having not engaged with technical courses for a while, I felt both overwhelmed and challenged. So, I turned to my trusted friend, ChatGPT.
Of course, I did not simply ask, “Please create a set of exercises for a database systems practicum.” That would have been too simplistic and would not have helped me learn much. My goal in consulting ChatGPT was to learn, with AI guiding me through the process. Additionally, I already had a solid foundation in database systems because, long ago, my undergraduate thesis was about DBS. A long time ago. At least, I wasn’t starting from zero, though I also wouldn’t become a hero. Maybe!
I shared how I truly had conversations with ChatGPT because I subscribed to it, which gave me access to the audio version. Every time I drove to campus, I would talk to it for at least 20 minutes. To me, it was like chatting with a good friend who was always helpful, non-judgmental, and endlessly patient. If I asked the same question 73 times, it would answer 73 times with a nurturing tone. Those conversations ultimately boosted my confidence in teaching and transformed my classroom into a dynamic theatrical performance—at least, that’s how I felt.
Next, Aji shared his experience as a researcher and vice-rector for research. Regarding AI, Aji stated that the question is no longer about whether we should use it but about how to maximize its effectiveness and efficiency. In his field, AI is indispensable. Aji even uses AI to develop new AI systems. Dina jokingly remarked, “It’s like AI breeding AI.” I felt the same way.
We both agreed that AI must be utilized in research because it allows us to accomplish many tasks more quickly and accurately. While Aji explained the technical aspects from an AI developer’s perspective, I complemented the discussion with general insights from a user’s standpoint. I thoroughly enjoyed the session, learning a lot from Aji and the participants’ responses.
As the discussion grew more intense and technical, Dina asked me to explain AI from a layperson’s perspective. I compared AI to a household assistant. When a new assistant arrives, they might not yet know the food preferences of family members. They need to be trained: the father likes tamarind soup, the mother prefers her eggs scrambled rather than fried, the child enjoys avocado juice with chocolate, and so on. Teaching a household assistant about family preferences is like training AI.
AI does not become intelligent instantly. It gets smarter with more training, just like a household assistant. Interestingly, AI generally does not forget what it has been taught, whereas a household assistant might. AI does not get tired, but a human assistant can become exhausted and emotional.
Regarding research, AI, in my view, is like a highly knowledgeable friend. If a person is lazy and dishonest, they might ask this intelligent friend to write their research proposal, conduct their study, and even write the final paper. They might have little to no involvement at all. And AI would never complain. It wouldn’t report or gossip about them later. AI is a loyal friend who never badmouths others. This reality presents a unique challenge, especially in the context of Ramadan—it becomes a test of a researcher’s integrity.
Technically, Aji demonstrated how Google’s NotebookLM can streamline many research processes. It can scan dozens or even hundreds of research papers in a short time and generate summaries. It can also produce analytical paragraphs or identify trends based on the papers it has reviewed. Additionally, it can suggest research gaps or potential topics after analyzing a vast amount of literature. Similar functions are available in other AI tools like ChatGPT, DeepSeek, and their counterparts.
Given AI’s capabilities, the real question is: How do we choose to use it? Will we use it as a learning companion that helps us understand concepts better, making us more knowledgeable and skilled? Or will we treat it merely as an assistant to handle tasks we refuse to do or understand? If we don’t understand the work, how can we evaluate its quality? More seriously, if we rely solely on AI for research, we might lose the ability to produce quality work without it. Imagine the dependency—that could be dangerous!
From my engaging discussion with Aji, we reaffirmed that researchers must still possess fundamental research skills. Identifying research gaps should be done by reading as many studies as possible. Comprehensive understanding must still be developed by summarizing, comparing, analyzing, and evaluating. These skills should never be forgotten by any researcher.
The good news is that AI can help us read faster, summarize more accurately, and compare information more efficiently. These accelerated processes enable researchers to work much faster. This is where AI serves as a reliable research assistant—but it will never be the lead researcher, as original ideas and creativity should come from the researcher.
In her book “Educating Humans with Machines,” Prof. Siti Murtiningsih argues that AI can support education. After my discussion with Aji at Arsana Space, I want to emphasize that researchers can collaborate with machines (AI) to create innovations and breakthroughs in research. Hopefully, this will be the case.
*Co-founder of Arsana Ecosystem (instagram.com/arsana.ecosystem)
PS: the article is translated into English from Bahasa Indonesia using ChatGPT with personal writing style of I Made Andi Arsana