How I Use AI to Enhance My Academic Career (Written By An Academic Physician)

Innovating Productivity in Academic Medicine

To ascend in your academic career, accumulating enough professional currency is essential. This currency in academia isn't just about conducting research but also involves teaching, participating in committees, and contributing to departmental growth. Each role, while contributing to career advancement, brings with it an inherent administrative burden that often detracts from the core scholarly activity: publication.

Consider the daily grind in academic medicine—how much time do you spend on activities that propel you toward your career goals? Administrative tasks, though necessary, often consume disproportionate amounts of time and energy, potentially stifling productivity and innovation. This is where Artificial Intelligence (AI) steps in as a transformative tool.

Since the advent of advanced AI tools like OpenAI's Large Language Model (LLM) ChatGPT, I have explored various ways AI can streamline the overwhelming administrative load we face in academia. AI makes mundane tasks easier, enhances decision-making processes, and improves creativity, allowing academics like myself to focus more on impactful scholarship and less on administrative minutiae.

In Slow Productivity, Cal Newport examines the shift from the Industrial Revolution's tangible productivity metrics, like bushels of apples picked or widgets made per hour, to today's less tangible knowledge work. He argues that despite the change in the nature of work, we still cling to outdated productivity measures. As a result, modern workers prioritize activities such as emails, meetings, and phone calls—actions that serve as proxies for productivity. Newport suggests that this approach stems from the difficulty in measuring the actual value of knowledge work, leading us to rely on these easily quantifiable but potentially misleading indicators.

Our work in academia can't exist in a vacuum. We need to take on roles within our organizations that might not perfectly align with our goals, though it helps serve a broader mission for the people we work with. So the challenge then is, how might we limit the impact of these activities on the three primary resources we have: time, energy, and money?

Acknowledging Newport's critique of our outdated productivity metrics, it becomes clear that academics are often bogged down by inefficiencies that traditional systems need to address. AI can play a transformative role, offering tools to redefine and streamline our approach to academic productivity.

The Reality of Academic Overhead

Your time and attention operate within a zero-sum game; every 'yes' to one task inevitably means a 'no' to another, a fundamental aspect of our finite capacities. In academia, as in all professions, specific tasks are necessary but may appear differently on your CV or performance review. These include organizing holiday parties or chairing committees, which are important for the organization but sometimes under-recognized in annual reviews or promotions.

The problem intensifies when your efficiency and reliability mark you as a go-to person for these tasks. Being recognized as dependable can lead to an increased load of such non-critical tasks. While this is a testament to your capabilities, it also presents a valuable opportunity to set boundaries and prioritize effectively. Fortunately, we have tools that can streamline these tasks, freeing us to focus more on what truly advances our careers and personal satisfaction.

I know many of you are unsure if you should use AI or how to do so best. That hesitancy makes sense. I want to offer another viewpoint and some ways you can start today.

Rethinking AI's Role in Academia

It's easy to view AI through the lens of a futurist, dreaming of its potential while overlooking its current impact. For the first time, we possess technology that augments our cognitive abilities, enhancing how we generate and hold ideas. In his book Co-Intelligence, Ethan Mollick presents compelling data on how pivotal technologies have boosted human productivity. The steam engine, for instance, increased productivity by about 20%. Estimates suggest AI boosts productivity by 20-80%—a broad range, indeed, still, it underscores AI's profound influence, positioning it as a modern counterpart to the steam engine, a cornerstone of historical progress. This isn't just potential; it's the reality of what AI is achieving for us today.

So what's the big deal? AI is more than just something that helps you brainstorm, edit an essay, or make pictures. Its effects are cumulative, meaning when you lean on AI to help jumpstart your work, you reduce the cognitive drag felt when starting or reiterating your work; its effects compound over time. It's essential to remember that this is the worst AI will ever be. It will only get better. How do we harness AI's strengths while preserving important human characteristics to produce original work?

Steve Jobs famously said:

"And that's what a computer is to me. What a computer is to me is it's the most remarkable tool that we've ever come up with, and it's the equivalent of a bicycle for our minds."

Jobs refers to how much more efficient a bicycle makes human locomotion than walking or running and how a computer does that for the mind. The bike makes the person no less human but is just a tool that allows them to get to where they're going quicker. AI is no different; a tool that, when used wisely, overcomes human limitations to help you get to where you're going quicker. 

There are valid concerns about AI and its place in generating art, music, and other creative outputs. How the AI models are trained has been rightfully questioned, as there are ethical questions around copyright and intellectual rights that we as a society need to sort out. The drive to produce new AI models as quickly as possible has led the major companies to make questionable choices to use data in unprecedented ways.

Just because AI is a complex, not-understood-by-many concept that will require asking hard questions of ourselves and society doesn't mean we should stick our heads in the sand or ignore it. This technology is going nowhere and will only become more integrated into our day-to-day lives. Considering how significantly AI will alter how we work, you need to begin exploring how the technology fits into what you do to discover its nuances, strengths, weaknesses, and ways to leverage it to help you do more of what you love. 

In the following sections, I will delve into specific instances where AI has significantly uplifted my academic productivity and career satisfaction, shedding light on the immediate benefits and long-term implications of integrating AI into academic practices.

Practical Applications of AI in My Career

The AI Award Nomination Author

One of my professional roles is to serve as the director of an awards and recognition committee. The committee aims to increase the overall visibility and recognition of the work done within our department, individually and departmentally. Even the smallest token of recognition goes a long way in promoting career satisfaction. 

Much of my role involves collating and organizing award opportunities at the institutional, regional, and national levels across numerous professional organizations and aligning them with faculty members in our department whose contributions match the award criteria. Each award typically requires assembling a nomination packet that includes the nominee's CV and a supportive nomination statement. This work is significant as it enhances our department's visibility and morale. While it may not be the cornerstone of my career, it is essential and rewarding. I strive to manage these responsibilities efficiently, ensuring they enrich our department and maintain my focus on other high-impact projects that also demand my attention.

I leverage AI to help with this role in two main ways: organizing data and drafting nomination statements.

Organizing Data

In this role, I streamline the management of award opportunities by leveraging Notion, an app equipped with built-in AI capabilities. I compile all award information into a comprehensive database, where the AI assists by summarizing each award's criteria and critical deadlines. This enhances the database's usability by making it easy to scan and allows for the creation of filtered views that simplify the process of delegating tasks to other committee members. This system ensures we efficiently organize and track award opportunities, maximizing our department's participation and recognition in relevant awards.

Drafting Nominations

Several months ago, OpenAI introduced CustomGPTs—configurable AI bots based on OpenAI's large language model. Think of these as small, task-specific apps with predefined prompts and actions. I leveraged this technology to create a CustomGPT that assists in drafting award nomination statements, addressing one of the primary barriers in submitting these nominations: the time required to craft compelling statements. High-quality nomination statements must reflect the specific contributions of the faculty member and how these contributions align with the award criteria.

The process begins by selecting candidates whose work directly pertains to the award's focus, ensuring each nomination is relevant. Subsequently, nominees provide a CV and a concise document—preferably a one-page summary—highlighting the relevant aspects of their work. This document does not need to be in prose; bullet points summarizing key contributions and their impacts are sufficient. This streamlined approach efficiently captures all relevant information, making the nomination process thorough and manageable. 

Having drawn inspiration from the blog Write With AI, I've built the Award and Recognition Assistant GPT from the following basic steps (reference Write With AI article that inspired this):

1. Establish the identity and role of the GPT.
2. Define the specific job of the GPT.
3. Create constraints for the GPT.

These steps help produce the best output from LLMs like ChatGPT. LLMs are just math machines; they combine words at random until some combination of words best matches a pattern learned through pre-training. That is, an LLM like ChatGPT is trained with (nearly) all of the available written content on the internet and, in doing so, has determined statistical probabilities that certain words and letters will appear together. The LLM doesn't attribute any particular meaning to the combination of words and letters it spits out. To help increase the chances it produces what you hope it will make, you need to give it some constraints and guidance. For the Award and Recognition Assistant, it starts like this:

"Remember your role. You are a world-class copywriting AI assistant specializing in drafting inspirational nomination letters. Your task is to create convincing nominations statements based on the information provided to you."

During the first step, I reiterate in the instructions that the primary goal of the GPT is to "create original nomination statements." 

I prompt the AI to create output from the framework of a copywriter, increasing the specificity of the work it will help generate. 

The second step is to give specific instructions for the actual task you want to complete. It's simple and to the point:

"Step 2: Review the nominee information uploaded with each task. Review their CV details plus any other information they have provided to get the full context and understanding of their professional accomplishments, scholarly productivity, unique attributes and experiences, and contributions to their field."

In leveraging AI to draft nomination statements, ensuring the accuracy and integrity of the information being processed is crucial. One known issue with generative AI, referred to as 'hallucinations,' involves the AI creating or inferring facts that aren't present in the source data. This can be particularly problematic in healthcare and academic research, where precision is paramount.

To mitigate this, I use a specific protocol when working with AI to prepare nomination packets. After collecting CVs and relevant summaries from nominees, I employ strict constraints within the prompts to reduce errors. For instance, any generated content not pulled directly from the input data is marked, ensuring placeholders are visible for manual verification. This step is crucial for maintaining the credibility of the nomination process and reflects our commitment to upholding the highest standards of accuracy.

I help to limit this by instructing the GPT in the following way: 

"Complete any tasks that are asked of you. If you have questions or need any additional context from the nominator, please ask."

"If you make up something about the nominee, put it in brackets [] so the nominator knows that this is intended to be a placeholder for them to replace later."

"...if you are not entirely sure, you can give the nominator multiple-choice questions to choose from to get additional clarity."

Again, reiterating the instruction in multiple ways decreases the probability that the AI will hallucinate and increases the transparency of the output to improve the "human-in-the-loop" effect that produces the best AI output. 

The last step I include in the CustomGPT is to reiterate to the AI that it can access necessary information in the documents I upload for each nomination statement while inviting the AI to ask questions if it doesn't have the information it needs. The upfront attention to proper constraints and guidance means less work on the backend to reiterate the work. 

In just a few minutes, you'll have a draft that is 80% there. The magic here is the reduced friction in going from nothing to the first draft, the most challenging part of any process. Since this writing style's main job is communicating someone's fit for an award, originality is not necessarily the goal. Instead, it references and describes factual information that matches the nominee's CV or other biographical information. With a bit of back-and-forth reiteration, you can dial in the tone and style of writing. 

The goal is not to craft something to win creative writing awards but to increase the overall number of faculty considered and awarded for their work. AI is a tool that helps get that done with less time and effort, so you can focus on other aspects of your job that require more creativity and problem-solving. 

Next, I'll show you how I use AI again to focus on communication, though this time with a goal of just-in-time information to foster collaboration with colleagues. 

The Paper Summarizer

Journal articles in academia are the principium of communicating your ideas, both directly when you author articles and indirectly when you find an article that describes a concept or approach to a problem you find salient for your challenge. I'll commonly share an article with colleagues I'm working with to share the insight gleaned from the paper. While there's a huge benefit in distilling and summarizing articles into your own words to improve your understanding and retention of the information, that's sometimes overkill and unnecessary when you need to get the gist across to someone with the least effort possible. 

Here's my workflow for sharing ideas and insights from journal articles:

I import each journal article I read as a PDF into Reader, a read-it-later app from Readwise. Readwise is a service that captures highlights and annotations from multiple places and imports them into a digital notetaking platform. I capture websites, YouTube videos, newsletters, academic journal articles, Kindle books, etc. I consume them this way. Given the breadth of content I consume, Readwise is a common meeting place for interesting information. 

Once I've reviewed the article in Reader, annotating the salient or interesting information, my notes are automatically exported to Reflect, a notetaking app focusing on "networked-thought" or backlinks that allow you to easily link ideas across your notes. The ability to backlink information in your notes enables you to create a Wikipedia-like structure that gives context to other information you may have previously collected or consumed. 

Reflect has AI built into the app; I use ChatGPT 4 directly from the AI palette within the notes, prompting it to summarize the notes as if it's an expert researcher, focusing on the essential ideas and optimizing for readability. The goal is to communicate the critical points of the research most simply. You want the summary to be quick to read and understand. I also prompt it to generate the three most essential take-aways at the end. You could do the same thing within ChatGPT or another AI. 

What's so powerful about doing this is that it isn't just feeding the paper into AI and having it spit out a generic summary. It first depends on you to process the information, annotate the ideas or concepts that you feel are most important, and then create a summary of those notes. You draw insights from the work and then use AI to help you recapitulate the information into a quickly understandable form. 

I like to take AI's output, create a shareable document, and embed the original paper. Typically, any note-taking app can share a copy of the note easily. I like Notion's publish feature, which creates a mini website. Another app I've begun to test is Craft. Here's a paper I recently shared via Craft: 

The Benefits of AI in Academic Life

Using AI in academic medicine decreases friction when going from idea to artifact. It's a matter of using AI to help you do more of what you love and less of what you don't. As I've experimented with AI across my work, I've found it to help jumpstart creativity and divergent thinking by no longer needing to start with a blank page. When I need to converge my thinking to refine a particular idea, it helps me summarize and communicate my thoughts. In a career where my time is not just constrained but also highly variable, leveraging AI helps sand the edges of the puzzle pieces so they fit together more easily. 

Over time, AI will become more technically functional and seamless in how it interfaces with our lives. As this happens, it will feel like a natural extension of our cognition. Instead of copying and pasting text into a separate app or web browser, your favorite AI will float over what you do, pulling together new connections, meaning, and insights without you needing to engineer the perfect prompt. The experience will be more about learning how to ask interesting questions from the AI and less about tinkering with the technology. As AI becomes more practical and less intrusive, your cognition will be freed to use toward more critical thinking and idea refinement. You'll be able to focus on the aspects of your work that your mind is best suited to tackle while leaving other parts for AI to handle.

AI as a Companion, Not a Competitor

The inevitability of AI in our world means you will either learn how to harness it or be replaced–whether by someone who has learned or the technology itself. We're currently at a point of inflection, an opportunity to begin exploring what's possible and learning how to use AI as a tool. Don't think of AI as your competition; think of it more as a catalyst available to anyone willing to understand its nuances and how to invite it into their work. Much of the attention paid to AI has understandably been about how it will affect our workforce, the types of jobs, and what kind of people it may replace. Indeed, this is important, and inevitably, the way people work will shift, but just like every other major technological breakthrough, we will adapt as it becomes integrated into our lives. 

A more helpful approach is to find ways to bring AI into your work now, to use it as a companion in what you do, and not think of it as future competition. The exciting thing about AI is that it is most helpful in improving peoples' weaknesses. Areas where it makes people more productive or creative occur where they have the least skill. This is good news because it means you can get the most significant boost by using it for things you probably don't like doing (because you're likely not very good at it!) Do you feel like you struggle to come up with fancy acronyms for your research? Have AI give you a list of 20 possibilities based on your study's name. Unsure how to generate a figure from your statistical software? As AI for the code to make it. Use the technology to help level the playing field, supporting you with the things you could use the most. 

Charting the Future: Embracing AI in Academic Medicine

As we stand on the precipice of a new era in academia, integrating AI into our daily practices isn't inevitable; it's imperative. The landscape of academic medicine is evolving rapidly, driven by technological advancements that promise to redefine what it means to be efficient, creative, and successful. The tools we ignore today will become the benchmarks of tomorrow's standards.

Embracing AI in academic settings is less about keeping up with trends and more about seizing the opportunity to shape the future of our field. The cost of innovation is substantially lower for the individual than for the organization, as Ethan Mollick points out. This disparity highlights a unique advantage—we, as individuals, can experiment and innovate with minimal risk but potentially transformative rewards. It is through our efforts that the collective advances.

So, as you reflect on your role in academic medicine, consider this: AI can be your most potent ally, helping to amplify your efforts, refine your ideas, and extend your reach. Whether streamlining administrative tasks, enhancing research capabilities, or fostering collaborative endeavors, the potential is boundless.

Now is the time to act. You can start small, perhaps with a tool that automates mundane tasks. Each step you take builds a bridge to a future where AI is not just a tool but a fundamental aspect of thinking, innovating, and achieving.

Don't wait for the path to be cleared by others. Chart your course, push the boundaries of what you believe is possible, and in doing so, inspire those around you to follow suit. By embracing AI now, you're not just preparing for the future—you're helping to create it.

Acknowledgement:

The author(s) would like to acknowledge the assistance of ChatGPT, a language model developed by OpenAI, for providing suggestions and feedback during the writing process. All work was read and edited by the author who accepts full responsibility for the work.