Blending AI Assistance and Original Thought: ISP-Myanmar’s Way of Data Management and Analysis

Photo – AFP

On August 15, 2026, ISP-Myanmar held its sixteenth 30 Minutes with the ISP event, titled “Blending AI Assistance and Original Thought: ISP-Myanmar’s Way of Data Management and Analysis.” This English translation of the event’s recap memo was published on September 3, 2026. The original Burmese version was published on September 4, 2026. This publication is part of research conducted under ISP Myanmar’s Socioeconomic Studies.

▪️Concept Note

ISP-Myanmar strives to unpack Myanmar’s shifting political and socioeconomic dynamics through empirical policy research and data-driven analysis. Yet, there are formidable data-gathering constraints. Conflict dynamics evolve rapidly, and access to information is severely restricted. While battlefield engagements and junta airstrikes have surged in recent years, verifiable information remains elusive. Furthermore, obtaining data on illicit economies tied to transnational crime and the conflict economy remains exceptionally difficult.

ISP-Myanmar relies entirely on Open-Source Intelligence (OSINT)—synthesizing credible media reports, official gazettes, stakeholder social media, and documentation from civil society and international organizations. Building on these sources, we address the research questions. However, systemic crackdowns on independent media, internet blackouts, aid suspensions, and shrinking resources continue to threaten public access to information and academic inquiry. Meanwhile, the rapid ascension of Artificial Intelligence (AI) is radically transforming the research landscape.

Zack Kass, an American writer, frames this technological shift as “The Next Renaissance.” During the early Renaissance (from the 14t^h to the 17^th century), the invention of the printing press around 1440 democratized knowledge, bringing art, philosophy, and science to the public through printed books. Today, artificial intelligence is driving a parallel paradigm shift. Knowledge that was once scarce, and  inaccessible now suddenly reached the masses. Modern AI architectures process vast datasets and execute complex analytical computations at speeds that far exceed human capacities and standard computing limits.

For data-driven research, AI’s utility is proving profound. It can now assist across the entire pipeline—from data gathering and analysis to drafting final reports. Tasks that once demanded grueling human effort can now be handled in moments. Yet, this speed demands equal caution. Ensuring data security, empirical accuracy, and adherence to Responsible AI principles remain non-negotiable. In Episode 16 of “30 Minutes with the ISP,” we pull back the curtain on our research engine. ISP-Myanmar’s young researchers share a behind-the-scenes look at how they track armed conflict, war economies, and human security—and how they balance artificial intelligence with human intuition and critical reasoning.

Welcome and warm greetings to all members of the Gabyin Community joining today’s “30 Minutes with ISP” program. I am Naing Min Khant, and I will be hosting today’s session.

Normally, this program features concise presentations and discussions on ISP’s research findings. This time, however, we will be presenting the process of our work. How we actually operate day-to-day and how we use AI technology.

To share why we took some particular approach in our work and why we organize this session, ISP-Myanmar’s emerging researchers and my colleagues Ko Zaw Htet, Ma Nann Kyi, and Ko Aye Chan will be leading the discussion.

After their presentations, we will address the questions and comments we received in advance from our Gabyin Community members, as well as take live questions and comments from those attending today. You can raise your questions and comments Live or via chat or Q&A buttons. Due to time constraints, if we are unable to get to all the questions during the live session, we will reply to them later via email.

Please keep in mind that questions and comments must not contain hate speech, insults, or derogatory remarks regarding race, religion, gender, sexual orientation, or differing political views. We have also posted these guidelines in the chat.

Today’s session is being broadcast live on the DVB TV News and ISP-Myanmar Facebook pages. We will now begin the discussion. Ko Zaw Htet, please start us off.

Thank you, Ko Naing Min Khant. Greetings to all our community members! First of all, I want to thank you for joining today’s session.

Following up on what Naing Min Khant mentioned, ever since we launched this ISP Gabyin Community, we have always wanted to do one thing. That is, we wanted to share how we conduct our research, in other words, our work processes, or what goes on ‘behind the scenes’. 

So, in this 16th episode of the “30 Minutes with ISP” event,  we want to share a key aspect of our work, specifically, how we gather and analyze data and the methodologies we use, even if we can’t cover everything, maybe, to show in part.

Back in 2024, we held a discussion in this program titled “Making Data Speak.” At that time, we discussed how we capture and present Myanmar’s conflict and social suffering.

This time, we want to connect our discussion to the unavoidable surge in AI technology, artificial intelligence. Regardless, AI has gradually integrated into our daily lives, making it essential for us to understand how to use it properly.

As we use AI more and more, we also risk becoming overly dependent on it, which can erode our independent thinking and human intelligence. At ISP-Myanmar, we also incorporate AI when gathering data and carrying out research tasks. When doing so, we constantly think about how to enhance and bring out our own human intelligence.

On the other hand, we must be careful to ensure that our writings, speeches, and publications do not sound robotic. In other words, we often say that what we create must reflect our own intellect and effort.

We plan to structure today’s discussion into three parts. Firstly on how we analyze research and data, the methodologies we use, and what constitutes our human intellect and effort. Secondly, how we integrate AI technology into our ongoing work. And thirdly, the precautions and best practices when combining AI with human intelligence.

Now, let’s start with the first part—what we do and how we do it. Aye Chan, please go ahead.

Thank you, Ko Zaw Htet. In my section, I would like to share what ISP-Myanmar is researching and how we conduct our research. I want to break this down into three stages. First, what are our primary areas of study, what does our research framework look like, and eventually, how do we put all these together to present our findings? 

Currently, ISP-Myanmar has five key research areas. The first is the Conflict, Peace, and Security Studies. The second is the China Studies Program, focusing primarily on Myanmar-China relations. The third is the Conflict Economy Studies. The fourth is Socioeconomic Studies and Climate Studies. The fifth is the Governance Studies, where we study Naypyitaw’s administration as well as the governance of various ethnic armed groups and other forces.

Moving to the second stage, I’d like to explain the criteria we use to select what to present to our audiences. We constantly strive to ensure our research, analyses, and assessments hit a convergence of three key criteria. We have briefly presented this during our “Data that Speaks” discussion in 2024 as well. We refer to this as our “Three Sweet Spots”. First, we assess whether it is current and relevant to Myanmar affairs. Rather than focusing on history, we want to discuss and find solutions to problems happening in the present moment. Second, We conduct our research using rigorous methodologies and examine our findings through strong theoretical lenses. Third, we try our best to provide policy relevant insights. What I mean by that is we try to make sure our data and insights can be useful for the policy decision-makers. So in summary, if the data and insights are relevant to Myanmar affairs, theoretically strong and methodologically engaged, and have policy implications, meaning  if it meets these 3 sweet spots, we proceed to  publish these for the public.

A vital part of these publications is data collection and analysis. So, for the third stage, I want to discuss how we actually collect the data required for this. There can be many methods for collecting such data. We rely primarily on open-source data which are publicly available and accessible sources. All the maps and analyses we currently publish are based on data obtained from these open-sources. Simply put, we present reliable data that everyone can find, access, and use. We also cross-check the data obtained from here using rigorous methodologies to ensure accuracy and reliability. So, the armed conflict data shown in the slide now, as well as rare-earth mining conditions and data on the location shifts of coal mining operations, all of these are built on open-source data. So these are a few sample analyses. What I have discussed just now is how we conduct research using our human intelligence. So, Ko Zaw Htet, please continue by discussing how we integrate AI to strengthen our human reasoning.

Thank you, Ko Aye Chan. Now, we want to continue discussing how we use AI technology in our daily operations and how we use it effectively. As Ko Aye Chan discussed in the previous section, ISP-Myanmar has five key research areas, and each has its own data collection processes. One thing we want to clarify here is that we ourselves are not AI technology experts. We are political and social science researchers. Rather than speaking deeply about technology, we simply want to share how it is used in research workflows like ours. 

So I will continue discussing how it is used in our data workflow. Broadly speaking, the data process can be discussed in three parts. The first part is collecting data, the second part is analyzing the gathered data, and the third part is creating figures or data visualizations based on this data so that the public can understand it and depending on the question we want to answer. 

So I will continue with the first part: data collection. ISP-Myanmar sources its data mainly on open-sourced data. AI is also extremely powerful at searching and gathering information from these open-sources. If a researcher can thoroughly and clearly explain their research question and background context to AI, it is extremely useful and can pull up a vast amount of information. Furthermore, processing numerical data and running analyses has become much easier. For example, almost all the office computer software we use daily, such as Microsoft Word, Excel, and PowerPoint, now includes AI features. Also, Google Workspace and Google AI have become very useful in our daily operations. Abilities such as summarizing large volumes of papers, transcribing audio files into text, brainstorming ideas for analysis, and organizing files systematically are immensely beneficial for research work and researchers alike. So Google’s NotebookLM, for instance, can be used extensively here. We also want to share another useful example. We search for information from open-sources like reports, newspapers, and photos. Previously, computers couldn’t read these. So we had to spend hours and hours re-entering data manually so the machine could understand and read it. In today’s AI era, this is easier, allowing us to extract data from photos and other information. 

Furthermore, at ISP, we use software like Flourish for graphic visualizations and QGIS for mapping and spatial analysis. These software applications themselves can now be used alongside AI. If you show a sample of what you want or how it was done previously and ask it to generate a design, it can start producing viable design concepts. Right now, AI companies are competing to build better services and capabilities. For us, besides Google’s Gemini, Microsoft’s Copilot, and Anthropic’s Claude, we also use other services. Lately, Claude’s capabilities have increased and become noticeably stronger. So if your prepared data is solid and you know precisely what you are looking for, no matter how complex or intricate the task is, AI can deliver results in a short moment. 

In summary, we integrate AI across data collection, data analysis, and data visualization as appropriate. There is no denying that advancing technology benefits research work. However, there are also matters that require great caution. Relying on AI from start to finish for an entire research workflow is completely inappropriate. So, Nann Kyi, please discuss what to be cautious about and how to balance artificial intelligence technology and human intelligence.

Thank you. To continue the balancing act of artificial intelligence and human intelligence, I would first like to share a few cautionary points about AI. First of all, although AI can collect and summarize information, it sometimes writes false information as if it were true. Because it gives out answers convincingly and neatly, we must evaluate its responses and outputs with great caution. The second point to consider is the knowledge limits of AI models. Because AI knowledge varies by version and often has time cutoffs, we must keep in mind that its knowledge can also be limited. This is because AI only knows the data provided by its developers and has no idea what is happening on the ground in real-time. 

Additionally, open sources and publicly accessible news sources often can contain inherent biases due to the nature of news reporting. To critically evaluate and discern such biases, we need the human intelligence of researchers. If we blindly rely on AI when analyzing biased content, we risk inheriting and perpetuating those biases, so this is an important point to watch out for. 

Moving on to the second part, I would like to discuss how we balance these challenges. For young researchers at ISP, when using AI in our research workflows and components, we follow guidelines. I would like to share three examples. First, humans must directly supervise the entire research process. Maintaining human supervision is crucial. When conducting research, humans must decide which research questions to answer and what hypotheses to make. Because AI-generated outputs in these areas can carry errors or biases, we must never take AI conclusions or interpretations without verification. As we often say at our office, we should treat AI at most like an intern, but maintaining personal supervision across all other processes comes first. 

Second is ensuring transparency. By transparency, we mean that if AI was used in research workflows, we must clearly and specifically document where it was used technically such as specifying which research section used which AI model and what prompts were applied. The third point is personal privacy and data security. For example, when conducting surveys, respondent data is sensitive research information that should never be uploaded or input into open-access or public AI tools. In summary, we balance AI and human intelligence at ISP by selectively using AI assistance to strengthen research methodology and theoretical foundations. However, assessing critical decisions such as alignment with the country’s current context, appropriateness, and relevance for the public and policy decision-makers relies primarily on researchers’ human intelligence and critical analysis.

This concludes our preliminary presentation on  ‘Balancing Artificial Intelligence and Human Intelligence: Data Research and Analysis Methodologies at ISP Myanmar’. We would now like to move to the Q&A section. We invite questions and comments from community members in attendance. You can ask questions and discuss using the raise hand button. We would like to start by inviting comments first. Are there any comments? Since no hands have been raised yet, let’s start with advanced questions from our Gabyin community. The first question asks: 

“It is said that AI can perform many parts of research. If so, what key qualifications do you think a researcher should possess over the next 5 to 10 years?” 

Could our colleague please answer this question?

I will answer this. First of all, thank you very much to those who submitted this question. This question is very good because it helps us prepare for the future and is a topic worth discussing broadly. However, I would like to share our view briefly here today. First, we assess that curiosity and the ability to ask good questions will become increasingly important. Second, we think critical thinking and analytical reasoning skills will become even more essential. Finally, we consider empathy and human compassion most crucial, and we believe they must be emphasized even more. This is because research is not just about numerical data; behind these numbers are human lives, social suffering, and challenges amidst conflict. In this AI era, while tasks can be solved easily and completed quickly with AI assistance, we must never forget humaneness, which we assess will become increasingly vital.

Thank you, Nann Kyi. I would like to share another question from the Gabyin community:

It was discussed that data is collected from open-sources. What kind of data does ‘open-sources’ refer to, and what does it include?

I will answer this question. This was covered in my section as well. When ISP-Myanmar conducts analysis and evaluations, we mainly rely on publicly accessible open-sources. Methodologically, open-sources can include many types of data, but I will provide a few examples here. Examples include reliable news media and their reports; reports and statements from government, non-governmental, and international organizations; official gazettes; and media posts from key stakeholders. Simply put, any publicly available source is considered open-source. Let me conclude my answer here.

Thank you, Ko Aye Chan. Here’s another advanced question, 

I regularly read and study the ISP-Myanmar’s publications. I am a student researcher myself. I am interested in how ISP creates its maps and whether it uses AI. When reading papers or holding seminars at the university, I reference ISP’s data, and I’d like to try doing this myself.

I’ll address this one. First, thank you for your interest in ISP-Myanmar’s publications. When creating maps or charts, we use foundational software through a step-by-step process. There are accessible, free tools suitable for students and the general public such as Datawrapper, Flourish, and QGIS. Do we use AI here? Yes, we do. We use it mainly for data preparation, conceptualizing design ideas, and brainstorming visual layouts. However, a unique aspect of ISP’s process is that we never use raw outputs directly whether from AI or software templates. Our ISP design team custom-builds everything from scratch to highlight ISP’s brand identity, relying on human creativity. For student researchers, tools like Datawrapper, Flourish, and QGIS are powerful and open-access options. Thank you.

Thank you, Zaw Htet. Now I would like to turn to today’s attendees. Are there any questions? You may raise your hand using the raise hand button. We have a raised hand from Myint Htun. You can unmute your microphone and ask your question.

Yes, my connection had a little lag, so I typed it in the chat too. The question might sound a bit silly, but roughly at what percentage or stage of the research process do you use AI assistance? It might depend on the research, but I would like to hear the team’s answer.

Yes, I will answer this question. To answer this, I would like to recall the three sweet spots research framework of ISP-Myanmar mentioned earlier by Ko Aye Chan. We aren’t saying our way is the absolute right way or that everyone must follow it, but we want to share our process. ISP-Myanmar has this framework for policy research. Among these three, we primarily use AI within the second part, to strengthen methodology and research process. It varies depending on the specific research project. AI is used only as a component within that specific section, while the other two parts rely entirely on human intelligence.

Thank you to Myint Htun for raising the question and Zaw Htet for discussing. Due to time limitations, we will end the Q&A segment here. Colleagues, please provide your closing remarks regarding today’s discussion.

To summarize today’s discussion, I would like to highlight three main points: First, ISP-Myanmar has a research framework aligned with three sweet spots, gathering necessary data from open-sources and evaluating it using human critical thinking. Second, AI technology has become an unavoidable reality in daily life. Rather than ignoring it, we integrate it into our daily workflows. Third, to prevent over-reliance on AI from dulling human intelligence, we actively balance AI and human intellect while following key ethical guidelines discussed today. What we shared today is a behind-the-scenes of how ISP-Myanmar conducts its research operations. We are not implying that our way is the only right way, but rather sharing our specific approach. We look forward to sharing more behind-the-scenes insights in future “30 Minutes with ISP” episodes. Thank you.

Thank you, Aye Chan. You can explore ISP-Myanmar’s research publications on our website at www.ispmyanmar.com and on our social media platforms. Thank you everyone who took the time to attend today. We conclude today’s program here.

▪️Appendix Questions

The question listed below was submitted via chat during the “30 Minutes with the ISP” event on August 15, 2026.

Which AI model is the best?

In a market driven by intense competition among global tech giants and major powers, declaring a single “best” AI is virtually impossible. No universal gold standard exists; suitability depends entirely on your specific use. For everyday tasks, OpenAI’s ChatGPT, Google’s Gemini, and Anthropic’s Claude offer versatile general performance. For research workflows, specialized tools like Google’s NotebookLM and Perplexity AI offer far greater utility. Rather than searching for an elusive “best” model, we recommend experimenting with various tools to identify their respective strengths and deploying them in combination.

Which AI tool is best suited for English-to-Myanmar translation?

No AI model is flawless enough to eliminate the need for human verification. Even in the current AI era, translating into Burmese remains exceptionally challenging due to ongoing limitations in Burmese machine translation data. With that said, practical testing indicates that Google’s AI models perform noticeably well with Burmese. ChatGPT also handles Burmese reasonably effectively. Rather than relying solely on a single platform, we recommend cross-testing different models and synthesizing the outputs to achieve your desired quality.