Interview by: Alperen Akbulut · 6 Oct 2026In this fourth installment of the Auditory Scene interview series, we host Asst. Prof. Atılım Atılgan of the Department of Audiology at İstanbul Medeniyet University to talk about the impact of artificial intelligence on audiology and academic work. His education began at Mersin University and continued with graduate studies at Gazi and Marmara Universities. We talked with him about AI entering clinical practice and hearing technology, the changing role of the audiologist, the possibilities in electrophysiological data analysis, his own daily workflow, and his advice for students and young researchers.
Professor, let us begin with a classic question that nonetheless matters a great deal to us. We know that your education began at Mersin University, deepened through graduate studies at Gazi and Marmara Universities, and that you now serve as a faculty member at İstanbul Medeniyet University. Over this long academic journey, how have you witnessed technology, and especially artificial intelligence (AI) in recent years, entering audiology and transforming our professional practice?
Getting involved with technology and technological tools, tinkering with them and trying to understand how they work has always been something I enjoy. That is why AI caught my interest from its very early days. At first I was using OpenAI tools such as the Playground. Back then there was not even a chat experience in today’s sense. Still, its help with text editing and translation impressed me a lot. At that time I did not foresee that AI would advance this far.
To be honest, I have not yet experienced AI much in clinical practice. In my view, audiologists largely continue to work the traditional way. Looking at the bigger picture, however, it has become possible to improve diagnostic algorithms in audiology by drawing on big data. Digital signal processing technologies in cochlear implants and hearing aids keep developing. These developments reach the user both at the basic hardware level and individually through smart learning.
The biggest transformation, though, is the one I strongly feel right now in my roles as an educator and clinical supervisor. Conveying hard-to-grasp concepts to students and running interactive, participatory classes has become much easier.
Let us continue with AI applications in the industry. We see that your work in the literature also includes studies on practitioners at hearing aid dispensing and fitting centers. Today, hearing aid and cochlear implant companies use machine learning algorithms very actively for noise reduction, sound environment recognition and device personalization. How do you assess the current state of AI integration in the industry, and do you think these autonomous technologies will weaken or further strengthen the audiologist’s role as the clinical decision-maker?
I believe the audiologist’s role will always continue. Our job is not only to adjust a device; in a sense it is about communicating and helping patients through their decision-making. However far technology advances, whether a patient benefits from a hearing aid cannot be explained by the device’s technical features alone. Yes, today’s hearing aids offer very important technological advances such as noise management, environment classification, automatic connectivity, the advantages of bilateral use and adaptation to different sound environments. All of these can positively affect the patient’s communication experience. Technology helps us use the existing hearing capacity as well as possible, but it does not eliminate hearing loss. Especially in noisy and difficult listening environments, no matter how advanced the device is, the person may still struggle to focus on listening and speech.
It is also not right to tie patient satisfaction to technology alone. Many factors need to be considered together: the degree of hearing loss, whether the person is ready to use a hearing aid, their expectations, their orientation, correct fitting of the device and the duration of use. That is why verification and validation of the hearing aid are so important. We can only understand whether the device’s features truly meet the patient’s needs through real-ear measurement, self-report questionnaires, speech-in-noise tests and other tools that assess auditory performance and patient satisfaction/benefit.
In short, I do not think the audiologist’s role will weaken; I think it will transform. As part of the technical work becomes automated, the audiologist’s role in clinical decision-making, counseling, communication, rehabilitation and adapting technology to real life will become even more important.
“I do not think the audiologist’s role will weaken; I think it will transform.”
Moving on to the academic and research side: AI has greatly expanded the limits of data analysis. You conduct research on complex topics that require intensive data analysis and precise physiological measurement, such as cortical auditory maturation in cochlear implant users and contralateral suppression of otoacoustic emissions in auditory neuropathy spectrum disorder. What new doors do AI models open for academics in processing and interpreting this kind of complex electrophysiological data?
I think AI has significant potential in the analysis of electrophysiological data. It can give us important support, especially in automated measurements. One of the main problems with electrophysiological data is separating the recorded response from noise/artifact and, in some cases, judging specific points on the waveforms by eye. Here AI/an algorithm can show us certain standards, patterns or averages. My approach, however, is that even if AI shows a result, the clinician should interpret that result in its context. How the test was set up, the patient’s condition at that moment, the test conditions, the artifacts that appeared and the clinical meaning of the result all have to be evaluated together.
In one of our studies, we examined how closely electrically evoked compound action potentials identified by an automated algorithm agree with detection by the human eye, and how much the automated algorithm could contribute to the clinician. Our study is currently under peer review. So my expectation is not that AI will remove the human from electrophysiological assessment, but that it will increase the human’s power of assessment.
We would also like to step into the “kitchen” and hear about your personal experience. Are there AI tools (ChatGPT, Elicit, Consensus, R- or Python-based AI-assisted coding tools, etc.) that you have built into your daily workflow when searching the literature, forming hypotheses or analyzing your academic data? How exactly do you use these tools to increase your productivity as an academic, and what practical tips would you give colleagues along the lines of “you must try this tool in this way”?
I use AI intensively in my academic and personal work. Where we are today, AI is not just a system we ask questions and get answers from. In the past, to reach a particular piece of information or expertise we always had to go through an intermediary, an expert or a particular institution. For example, to get an article you had to go to the library, get help from the people providing that service and spend time, and sometimes money, on it. Today, with a single query we can reach a large amount of information in a very short time, have our files analyzed and work on them.
I use more than one system, and each has a different area where it is strong. I use a reference management application called Paperpile. This tool stores full texts in the cloud. I installed the open-source Gemma 4 model on my local computer. Through a setup I use, this local model automatically generates a Turkish summary of every new article I add. This is very valuable to me, because instead of reading hundreds of articles in full at once, I can more easily identify first which articles are really relevant to my work. I can then import the full texts of the articles I select into Gemini Notebook and search only within those articles. This makes synthesizing and understanding the information and generating new ideas much easier. I use Claude mostly for agent-based and code-based work. There I can combine different skills to build more complex workflows.
For a large share of my everyday, quick tasks I use Google Gemini. Because it works connected to the Google ecosystem and interacts with the calendar and other tools, it is very practical in my daily work.
ChatGPT has a separate place for me. The fact that voice interaction and conversation-based working have become so advanced recently matters to me, because I have largely stopped typing on a keyboard. I speak, dictate, think, and by talking with the AI I turn all this into documents.
What really matters to me is not a “give data, get output” system. For example, saying “prepare an Excel file for me” and using the resulting file as it is, is not how I work. I work more like “go there, find this data, combine it with that information, then add it to the document in that folder.” In other words, AI has become less a content generator for me and more a working system and academic assistant.
Beyond that, I use it when preparing lectures, developing applications, building websites, in my music-related projects and even in administrative work. For example, I redeveloped music projects I had done in the past with the help of AI. I even built a personal mobile app from a congress program.
I am wary of using AI output as it is. I always review it, add my own knowledge and thinking, and only then use it.
My main advice to colleagues is this: use an AI tool not just to chat and use its output directly, but to broaden your knowledge and perspective. And turn the tasks you do over and over into a system.
The tools Atılım Atılgan uses, as he describes them.
As a professor in direct and intensive contact with undergraduate students at İstanbul Medeniyet University, you closely observe the new generation’s relationship with technology. Audiology undergraduates may sometimes tend to use AI at a shallow level, just to prepare assignments or produce quick summaries. So that they become the “digital health professionals” of the future, how would you advise your students to set up AI as a tool for professional development and learning throughout their undergraduate years?
I do not ask students for assignments made of long texts, because a student can get that information from AI in a few seconds if they want. So what we need to measure should no longer be only “do they know the information?” We should focus more on how they use the information, how they interpret it and how they solve a problem.
What I want more is for them to understand and to apply. For example, I find case assignments and simulations more valuable. I think the simulations students do before they meet patients are very valuable for learning. In audiology especially, being able to experience concepts students find hard to understand, such as masking, through AI-assisted applications and simulations makes learning considerably easier. That is why I develop various applications with AI tools and share them with my students. For example, I made a masking application I developed available to my students through my website.
When I ask students to prepare a research proposal using AI, I also ask them to declare how they used AI.
I have also changed my approach to course materials. We used to give students mostly slides. Now, before class, I give my students more comprehensive texts in the form of book chapters, built from notes and graphics I prepared myself. When they come to class, students do not just listen to me; they run simulations, take part in group work, evaluate each other’s work, discuss and practice. I think what will matter in education in the future is not that students get more information from AI, but that they learn how to use the information they get. Reaching information is very easy today. The real issue is being able to ask the right question, evaluate the answer and apply it to a real problem.
Concretely, I recommend three things to students. First explain the topic in your own words, then ask AI to find what you are missing; not the other way round. Use AI for patient simulations. When studying with AI, give it notes, a book chapter or an article whose accuracy you are sure of and ask questions about the topic. Ask it to teach you the concepts you do not understand. Always compare what AI says with the textbook, because without the basic knowledge you cannot catch its mistakes.
Always compare what AI says with the textbook, because without the basic knowledge you cannot catch its mistakes.
“The real issue is being able to ask the right question, evaluate the answer and apply it to a real problem.”
Since we are talking about students’ view of AI, let us turn this into a more concrete “toolkit” suggestion. For undergraduates who are just learning the clinical and theoretical foundations, and for graduate students doing in-depth literature searches, article reading and data analysis, which specific AI tools (for example for literature searches, language editing or research planning) would you recommend? What are your favorite AI platforms or plug-ins that would make students’ academic lives easier?
Rather than recommending a single tool, I think it is better for students to learn to use several tools together. For undergraduates I would first recommend Gemini Notebook and a general-purpose language model (for example ChatGPT, Claude or Gemini). But they need to be used for different purposes. You can use Gemini Notebook to create reliable study materials from sources you choose. You can use general-purpose models to have a topic explained in a different way, create cases, test yourself or discuss a topic.
Finally, we would like to close with a word for young researchers continuing graduate studies or at the thesis stage. Getting AI support for statistical analysis, literature synthesis or academic writing in English speeds master’s and doctoral students up a great deal, but it also brings serious ethical debates and risks of “hallucination” (generating false information). How can young researchers make the most efficient and safe use of AI assistants in their thesis and article work without crossing the boundaries of research ethics?
The ethical use of AI tools in scientific research is very important. Even if AI produces a completely correct text for us, using it without human contribution is not considered ethical. You can ask AI tools for information and use them to enrich a text, but you need to declare how they were used. Systematic and efficient work, from the start of a study to publication, has become easier with AI tools. In this sense, using these tools with an ethical mindset requires a somewhat more systematic structure. I think we should keep searching the literature the traditional way and use automated search tools for detailed searches. However good AI academic search tools are, we should never use the texts they produce without getting to the article itself. Gemini Notebook is really useful here; when you use it, it lowers the chance of making citation errors.
I find articles through traditional searching. Then I continue with AI-assisted searching using deep research and MCPs such as PubMed and Consensus. I then add the full texts of the articles I find to my reference system. The articles are summarized automatically from the full text. I access these summaries through Claude Cowork and identify the articles most needed for my research. When I get to the writing stage, Gemini Notebook comes in. I import the full texts of the articles into that application. That is when the chance of citation errors drops very, very low.
The steps Atılım Atılgan describes, in brief.
He never uses AI output as it is at any step: he always reviews it and adds his own knowledge and thinking.
As for data analysis, using tools such as R and MATLAB has become more accessible with AI support. With very little coding or software knowledge, it has become possible to run in-depth analyses and create impressive graphs. Here too, it is important to verify, checking the data with other tools. I think a researcher should be able to do the statistical analysis of their own research. Before AI, statistical consulting was often used in theses. Now it is as if AI has taken over some of that consulting. I think, apart from advanced statistics beyond their knowledge, a researcher should be able to manage the whole analysis process themselves. From this point of view, verifying the statistical support provided by AI by hand is very important.
Thank you. Having this conversation was a real pleasure for me as well. Alperen and I have a friendship that goes back to our days at Marmara University. I have long followed his interest in AI and technology, and especially the work he does in this area. I wish you every success in your publishing.
“Even if AI produces a completely correct text for us, using it without human contribution is not considered ethical.”
In this fourth Auditory Scene interview, we talked with Atılım Atılgan about the impact of artificial intelligence on audiology practice, education and academic research. The takeaway is clear: AI does not replace the audiologist or the researcher; used well, it strengthens their judgment. The condition is to question the output, go back to the source and declare its use openly. See you in our future interviews.
Auditory Scene