AI-Driven Research Transformation in Bangladesh: Machine Learning and Data Analytics. USETDA 2025.

Hello everyone and welcome to session 14. Uh we're going to talk about AIdriven research transformation in Bangladesh. A study on the integration of machine learning and data analytics in scientific discoveries. Sharing this information with you is Dr. uh Muhammad Zilla Raman who is a librarian at Ansula University of Science and Technology in Bangladesh. My name is John Hagen and I will be your moderator for the next 50 minutes. Um we'll reserve we will reserve five to 10 minutes at the end of the session for discussion. Um and a quick reminder before we begin um during the presentation portion um participants should keep your um audio muted but please feel free to use the Zoom chat tab to ask questions which will be addressed during the discussion segment. Um thank you all for joining and I'll turn it over to Dr. Ramen. Uh thank you John for a nice introduction introduction of this session and uh I would like to welcome everyone uh to uh this uh this session uh the breakout session 14. Uh my research uh topic is uh AI driven uh research transformation in Bangladesh a study on the integration of machine learning data anal analytics in scientific discoveries. Hello, warm greeting to all present as we gather here today. I am excited to introduce my research on AI drive and drive and research transformation in Bangladesh. A study on integration of machine learning and data analytics in scientific discoveries. uh today's discussion uh yeah the overview of the discussion that is uh the background of the study literature review methodology findings and result discussion conclusion and recommendation uh AI artificial intelligence uh a widely uh discuss discussed uh term in not only in scientific ific arena but also in academia and industry also. AI particularly machine learning uh machine learning and data analytics is transforming global research practices. uh enables proc uh processing of vast data set uncovering pattern and accelerating discoveries widely applied in deployed nations across sector like medicine, climate science and industry. In particularly the Bangladesh adaptation in research is still in very early stage. Emerging application is key sector such as healthcare, agriculture and environmental sciences. Uh for example includes AI based diagnostics and and appreciation agriculture and tools showing early benefits. What are the challenges in Bangladesh? Limited conceptual infra infrastructure, shortage of skilled professional in AI and data sciences, restricted access to highly uh high quality research grid uh data sets. Why this study was conducted? The aim of the study is to assess AI potential in overcoming research limitation in Bangladesh. The objective of this uh of the topic that is they evaluate the current level of AI adoption in research. Identify major barrier to widespread AI implementation. Proposed strategies for faster and AI supportive research ecosystem and what border context. The study position Bangladesh within a global AI research movement highlights both countries untapped potential and the urgent need for policy and institutional support. uh in literature review actually the literatur was conducted based on four uh major um thematic area that is AI and the scientific discoveries worldwide machine learning applications in Bangladesh and in research data analytics in research transformation challenges in AI adoption in developing nations and from the first of the discussion The findings that is AI the that is the machine climate science and geo geonomics application include protein protein structure prediction and drug discovery that disaster forecasting research learning application is research machine language supports pattern recognition prediction and automation in diverse field in healthcare. It improves diagnostics in agriculture in enhance propion and past detection in data analytics in trans research transformation. Data analytics enables larger skill data processing identify trend and guiding evidence-based decision. It strengths strengthens the geo economics and uh social sciences and interdisciplinary research and challenges in AI adoption in developing nations. Barriers include weak infrastructure, lack of expertise, high cost and inadequate policy framework. Actually in Bangladesh AI use is growing but hindered by limited resources and training. Uh this resource uses two method of uh one is qualitative method and another is quantitative method. Um sorry indilize structured survey to gather uh measurable data data collection as was convey uh was a survey distributed to academic resources research institution and AI professionals in Bangladesh designed to access the content of the AI adaption ac across various research field collection data on challenges faced uh when integrating AI technologies and data analysis. A descriptive and inferential statistics to analyze survey responses evaluated and impact AI technology on research productivity using statistical method and in similarly in quantitative method it was data collection uh case studies semistructured interview data analysis use qualitative analysis technique such as interpret interview intercripts uh prescription and attitude and experiences stakeholders regarding a driving research and identify trend barrier and opportunities in AI transformation uh infrastructure and expertise and what is the uh finding uh and results of uh this investigation actually adoption of AI in Bangladesh Bangladesh research uh uh which is gradually increased Over the past 5 years, adoption remains limited to a few discipline and is mostly project based. Growth driven by collaboration with international partners and donor funded initiatives. Overall use of AI is still modest compared to global standard due to resource limitations. Actually I identified the three areas that AI is widely used in Bangladesh. One is agriculture. Two is healthcare and the environment. Uh in agriculture, AI used to predict crop and and crop yields, detect paste uh controls and irrigation contributes to addressing food security challenges agriculture dependent economy. Healthcare AI drive and diagnostic tools improve speed and accuracy to enhance the investigation. They enhance the effectiveness of medical survey and environment AI applied in climate data analysis, flood prediction and air quality monitoring uh and support climate resilience in country and environmental risk and recently I can give you an example of that this environmental impact that the Bangladesh now the uh uh the ministry of uh environment and the climate they are every day they are analyzing the air quality of different districts and it describes uh it announces through the uh broadcasting finding result that is the institution and infrastructural barriers Limited conceptional infrastructure and research labs, poor access to highquality data set, inadequate funding for AI research, shortage of trained professionals in AI and data science. These factors slow the development of a robust AI research ecosystem. And what are the stakeholders perception and attitude toward AI? researchers, policy makers and industry representatives are cautiously opt optimistic about AI integration. AI uh recognizes potential to address national challenges in health, food security and climate change. Concerns raised about sustainability, ethical risk and financial cost and the emphasize on the need for stronger collaboration among academia, government and industry, policy support and targeted investment, capacity building and AI education initiative. Uh interprets some of key findings. AI adoption in Bangladesh research is progressing but remains uneven across sectors. Early implementation are seen in agriculture, healthcare and environmental research. Broader adoption is limited by infrastructure and institutional barriers. Findings highlights both the potential and challenges of AI in developing country context. What are the opportunity for AIdriven research in Bangladesh? For agriculture use the predict analysis analytics for yield optimization and climate resilience. Healthcare is used AI assisted diagnostic can is pressure on over the medical system in environment. AI tools can enhance environmental monitoring and support climate adoption strategies. These sectors offer high impact opportunities for a scalable AI integration. compar comparative perspective with other countries. Bangladeshi adoption is still nent compared to developed countries like USA and China. Developed nation benefits from robust government policies, a strong industry, academia partnership, advanced concept, comput competition infrastructure. Bangladesh shares a similar growth pattern with other emerging economics indicating gradual pro progress emphasize the risk of widening global digital divide if strategic interventions are not made. Policy and institutional implications argent need for policy reform and institutional development. The key recommendation are invest in high performance computing infrastructure, develop an integrate AI focused academic curricula, a strengthening collaborating among the academic industry and government. A coordinated policy framework can boost national research productivity. Position Bangladesh as a regional leader in AIdriven research and innovation. Uh now I am almost to uh conclude my presentation and the recommendation will be done. The summary of key insights. This study underscores that while Bangladesh has made initially a stage in adopting artificial intelligence for research, it application remains uneven across sectors. Agriculture, healthcare and environmental science stand out as promising field where AI has already been began to generate measurable benefit. However, infrastructural gap, shortage of expertise and limited funding continues to restrict widespread adoption. Overcoming these various requires deliberate policy and institutional action, limited access to high performance computing facilities and quality data set remains a significant constraint. Addressing this requires investment in digital infrastructure, strengthening data governance and promoting open data sharing framework. Strategy for strengthening ecosystem. Building a sustainable ecosystem will demand coordinated effort across government, acade academy and industry. Introducing AI focused curricula in higher education. Offering training program and incentivizing public private partnership can help build human capital and innovation capacity. Furthermore, adapt adopting clear regularity uh a regulatory guidance and ensure ethical AI practice while fostering trust among stakeholders. Further research should explore a sector specific EI application particularly in climate adoption, disaster management and public health area highly relevant to Bangladesh development priorities. Comparatively studied with other emerging economics could also provide valuable lessons on scaling integration and resource constraint context by addressing existing challenges. and embracing uh innovation, Bangladesh can position itself as a regional leader in AIdriven scientific discoveries. Uh thank you all. This is the end of my uh presentation John. >> Well, yes, thank you sir. Dr. Ramen, thank you so much. It was very enlightening to learn about the landscape of AIM in integration in the life in Bangladesh. So my question to you maybe just to start the process going here and the rest of you can think of maybe some questions you'd like to ask Dr. ramen but in terms of the impact of artificial intelligence on your students and in particular um in the research or um let's say the literature review process when they start to write their thesis and dissertations. I know earlier in your other presentation you spoke about using um for example um turn it in as a way to detect plagiarism and copyright issues and so forth. But in terms of the research process itself, um for example, are your research librarians well-versed in various um um prompt engineering and various um AI applications um in which they could um help their students search um for what research has been done previously in their area and um you know help determine what area of focus they will do with in writing their thesis or dissertation. Um I hope that was clear enough for you to understand. Uh yes. uh in for this particular research actually uh uh I declared uh when I submitted Abstract to the ETD 2025 and my declaration was that this paper uh uh to generate or develop this paper AI was used to generate uh literature review and uh uh to generate some uh context to or get some idea insight into the uh uh the the problem because you know as we are a library and information science professional actually we have a little idea about the healthcare what is going on and the agriculture system and the production system and the environmental impact that is going on and sorry and uh to get some idea about the all these the progressive field that has been using uh AI driving research already conducted and they are getting benefit from the AI AI uses in different field. uh I used AI uh to generate the idea in to to get a a preliminary idea about the uh the topic and uh when I got the idea and then I started developing the other resources and I contacted the literacy review and the literacy review basically was done on based on the four focused areas as I I earlier said uh it was uh uh AI and scientific uh discoveries worldwide, machine learning application in research, AI data analytics in research transformation and challenges in AI adoption in developing measures and uh from these four area literacy review I I I got a greater insight into the uh topic that the challenges in AI adoption in developing nations that you you you can see that barrier include weak infrastructure, a lack of expertise, high cost and inadequate policy framework and and and like other developing countries in Bangladesh, it is uh growing but hindered by the limited resources and training and uh uh the in in my previous presentation I told that the AI basically in Bangladesh and around the world We uh uh the the topic is spreaded and and we hear this topic that AI is coming uh though the the research has been conducting last two decades. one AI developing AI artificial intelligence and application in the human life and in various field and uh finally it came out to our existence in last uh uh last few years ago and uh the its uses and and and uh widespread use actually depends on many things many factors and uh uh as you know that uh if there is No lacking in expertise and uh if it shares the cost and inade uh the country has sufficient uh uh policy framework uh then uh definitely AI uses in every field uh uh it will uh enhance day by day by day and in the universities uh you know uh as I earlier said that the turnitin is being used to uh used uh to u take the pagerism in research and uh and uh and uh in in submitting thesis and the the jury board they are evaluating the research the report that uh the through the turn it orism checker tools uh presents while uh uh conducting a research Particularly in my uh in this this this research I used uh uh various uh method including qualitative and quantitative and uh uh you know uh in our qualitative research the uh application of quantitative uh approach uh is uh very difficult to analyze is uh uh the and at the pres data uh but in some cases uh uh some measures uh can be made with the uh quantitative data. Furthermore, the data analytics can trick were employed to analyze survey response and interview uh transcript and providing insights into the pattern and trend in adoption and descriptive and inferential statistics were used to evaluate the impact of AI technologies on research productivity. where uh while qualitative analysis also utilized to use the uh uh uh perception and attitude of stakeholders toward AI driving research. John >> um well thank you that was um a fairly comprehensive approach. Um so it's interesting to know that your um the paper actually this presentation and the paper underlying paper um that you wrote um you used AI to help structure and also to do the basic literature review um in preparing um to gather knowledge about this topic. Um and I suppose this would be an appropriate approach for you to for example instruct a graduate student um when they were beginning to do their research um you know uh for a thesis or dissertation. >> Yes, of course. uh uh when we instruct our students to do research and uh we strictly follow the research integrity in uh in the research uh in uh perspective of my university that's university as I told earlier that uh we we uh in we we have developed a policy uh uh to uh to check the research integrity in uh by using the uh using uh uh the tool for for for detecting the similarity that they use in their thesis and dissertation and u uh for for example when uh my the hard copy thesis the hard the traditional thesis one when I I started conversion into uh the digital world. Then my administration told me that uh it should not be because there there may be some content that used uh uh without acknowledging the uh proper sourcing or uh proper acknowledgement to the uh uh proper credit to the the original author. Then uh it it stopped. But recently the thesis that I we have been uh receiving uh in in my library strictly followed the context of uh AI uh uh AI detection tools and uh turnitin is quite enough to detect AI uh generated content and uh when it seems that this is the the beyond the tolerancy And that we uh have set a threshold that uh this amount of uh AI content will be allowed or this amount of similarity will be uh uh tolerated uh u as a threshold uh to particular thesis or dissertation. uh uh it it varies uh uh from across the departments that is the science faculty they're using this uh there are some uh standard and the earth faculty and arts and humanities they also uh they set a benchmark on getting the AI drive and the similarity index. So uh uh we we have been started the using of AI and AI related uh policies and uh uh in in uh not only in uh the ETDs and uh but also in project and recess also journal. >> Great. Well that was a very wonderful um uh reply. Now um I was just thinking of the earlier um presentation. It was a the the workshop rather on the system called illicits and they um allowed them to um to evaluate um various um AI systems the effectiveness and so forth and they creating policy recommendations basically for their campuses. So, I would point you to the University of Iowa for um some excellent resources if you're looking for something to implement on your campus. Um so, I see that Jenny uh from Virginia Tech Libraries has her video panel open. Um Jenny, would you like to um tell us something really interesting? >> Well, I um I I was actually interested in the same thing you asked about. So, thank you for um Dr. Ramen for your uh responses on those questions. Um I was also wondering um what you might have mentioned it but I didn't quite catch it but uh if uh there were particular tools you like using right now for this purpose like creating a literature review or uh exploring the context of a new topic and also if there are particular tools that it seems like um you know you were exploring use of AI in in Bangladesh in research. if you noticed uh there was like a lot of people using particular tools or if it was quite varied um or if there are like tools specific to Bangladesh that um maybe have data sets focused on Bangladesh context and work or other data particular to to to the region. um that you think would be interesting for for people out in other countries to try? Uh thank you Jenny for your uh interesting question and uh if I understand from your question that uh actually we uh our university student and uh we use chat GPT and uh uh you know chat GPT has released uh uh many updated version and uh uh our students they used this GP GPT to generate their it is review and they conduct uh uh sometimes I found that in my uh digital lab in um in in my library that they're using JPT and other popular uh AI tool uh and also they sometimes use deepseek and uh I know there are many other uh AI tool currently being used in academia and industry and my students particularly they use CAG GPT and uh uh Bangladesh has no this kind of bigger data set and uh that being used for uh developing or writing or uh analyzing um the Bangladeshi based the research uh right now actually we share uh the the research that have been conducted and published the article around the world by the Bangladeshi authors. And uh our students they uh sometimes use uh AI to like chat GPT release 4 and five or CGP plus and other versions. They they sometimes use uh uh for uh developing their codes and in pro programming languages and uh they are effectively and very uh successfully using all these tools uh in their um in their academic activities and our university uh sometimes provide uh uh provide training to the students. on AI through the institutional quality quality assurance test which is popularly called IQC cell and the director of the IQC cell. They they uh frequently conduct uh uh uh one AI uh how to use and what it would be and what are the uh ethical aspects and ethical perspectives of AI in research and academia. Uh, Jenny, >> thank you very much. >> Jenny, I was going to ask you um what your particular favorite AI tools or resources are. And maybe it differs depending on the task. I don't know, but um I would be curious to know what your favorites are. >> Um well, I'm I'm still I don't know that I have a favorite yet. I'm still trying out most of them. Um I've used chat g chat gpt a good amount. Uh we had a structured pilot. Um and I don't think our institutions made a definite decision yet about whether to move forward with an institutional license. We we had a temp a short-term license in order to try it out. Uh but we also had a pilot of Microsoft co-pilot um which I also participated in and we have as an institution licensed that. So >> you do have to pay for it um from your department funds. Uh but the benefits of the institutional license are that it's not training on your prompts. It's not training on your data. um and what you share is not going to be you know made part of its public data set or incorporated in any way. So I guess the benefits with that are the Microsoft copilot uh is embedded within if you have the full licensed product it's embedded within each uh tool. Um, so you have kind of a co-pilot for my for PowerPoint, a co-pilot for um Outlook, for Excel, for each application, and then 365 in some ways goes across applications. I think in my imagination, it would have been able to do more things across applications than it is currently able to do. Maybe at some point it will be developed enough to really like for example if you wanted it to create a schedule for you um that you could share with other people in a like you know nice visual way you might be able to say in Word um or in Microsoft Copilot please go look at like these calendars and put together a schedule in Word and it doesn't quite work that well. Um but you know the yeah using it for um some data exploration uh for um outlining slides or adding visuals to slides sometimes depending on uh the visuals it creates. Uh those are some some components. Chat GPT builds you. I've you I tried out Microsoft Copilot 2, but chat GPT did seem a little more user friendly in terms of exploring new topics and was easier to get it to site its sources um more effectively. I haven't done a lot with explicit. I've tried consensus which I liked um because it gave uh kind of some context around each source that it provided and it seemed pretty reliable in that the sources it cited were real and you could um look you could uh you know easily click on it and learn more or you could you know pull them together and get more summarized analysis from it. But I don't have the paid versions of either of those. So, >> excellent. Excellent. uh actually I didn't use co-pilot of uh yeah Microsoft in my country uh and my students they are using basically JGBT and uh deepseek and uh some other popular tools and you know uh after uh all these sources are not free uh for the students. If in free version if you conduct few uh few action uh then it becomes please update the upper version or something like this kind of message comes and if you analyze a bigger thing or if you call to generate uh uh uh some images or some graphs then it uh reminds you to update in uh the the uh the version that is uh which is uh the purchased only. So it continuously divert you to uh purchase the tool and which is sometimes difficult to pay uh money from Bangladesh by using credit card or something uh so very difficult. Uh do you uh does anyone have any experiences using Google Gemini? That's become one of my favorites just because it be it's become so ubiquitous in the browser. um that I um I just enjoy it and um it seems to be fairly accurate in terms of results and I've actually put together a um kind of a position paper. We're looking at a potential merger with organizations and that helped me gather my thoughts actually. So I found it very useful but I'd be curious to know if anybody else has a any experience with the the Google Gemini. Uh I guess not. Okay. Well, there is another area to explore definitely. >> One minute is remained to lift the session. >> I'm sorry, Dr. Rama, can you repeat that? >> Yeah, it almost close to the end of this session. >> Oh, yeah. We have another Yeah, I mean, we have a nine minutes. I'm going to leave it available if anybody wants to have any parting comments or thoughts. Um, another thing I was going to mention too is it largely depends on the applications um that we use. Now, for example, I mentioned um and maybe it was in another session, but about using Adobe um Express and our nonprofit USCD has has access to that um for free because we're a nonprofit and we applied for that. Um but it has very extensive um applications to create um illustrations and graphics and so forth. Um, however, I thought it was interesting to note that when I went to um upload a graphic um to Adobe Express and then it asked if I um owned the copyright um to the actual image and if I did not, if I answered no, it would not let me upload that to the content. So, obviously Adobe is aware maybe there are some liability concerns. Maybe it's also to help the end user to train them to guide them to use appropriate sources and so forth. But then I went to do the same request um on sh excuse me on Google Gemini with a basic text prompt and an image that I uploaded and I wanted to to create a similar but different image and it did it automatically without regard to copyright. So, uh different approaches um to that. But I think it is interesting that at least some of the AI um you know manufacturers or whatever you want to call them uh creators are um aware of these copyright issues and at least trying to have some sort of um I don't know a gateway or some sort of at least you know pass through before they just ingest things willy-nilly. The other thing I thought was interesting that Jenny you mentioned earlier that is you pay to play. So of course if you want free access then they'll say that's fine. You just have to give us all the data. Let us keep everything that you stuff into our system when you're doing your queries. Now if you do your institutional paid you know um access and apparently they will let you opt out of that and so they don't then ingest all of your information and that then becomes fodder for other AI creators who want to steal your information. Um it's a very wild wild west sort of situation but I find it fascinating nonetheless. Yeah, wonderful experience that John you shared and uh I would like to thank you everyone for uh participating in this uh fruitful discussion and uh uh I would like to thank the uh ETD, NTLDD, US ETD and the particularly John because the US ETD uh enjoying its uh 15th anniversary uh this year and uh I have been participating USD uh last couple of years and uh I I have been gathering vast uh knowledge from the discussion and I am really grateful to every uh their discussant author and the scholars they who are participating in ETD and I wish this uh ATG 2025 a great and a big successful one. Well, thank you so much Dr. Ramen and we always find your um presentations very fascinating and not only in terms of the things going on in Bangladesh but obviously the implications for the rest of us in the world and how that information trickles down to to you all to you all in your part of the world. Um, just a quick parting comment. Um, Jenny put in the com the chat tab here. Um, she says, "We're not we're actually not allowed to use deepseek at West Virginia Tech, pardon me. Um, for Virginia Tech related activities." So certainly there's some campus policies and some sensibility I suppose. And then um, uh, also the comment um, my colleague Dr. Silver Dr. Sarah over with the libraries at Virginia Tech provides a workshop on AI tools for research. Um if it's of any interest um the link is in the chat here um so you can um seek additional resources. So again thank you so much everyone for joining us and thank you Dr. Ramen for your enlightenment and um so we'll conclude the session uh and um let's please thank uh Dr. Ramen. Thank you.