Abstract
Experimental design is a core competency of scientific training with profound implications for research rigor and reproducibility. Yet, trainees often receive minimal guidance to structure their thinking around experimental design. Metacognition—reflecting on one's own thinking—offers a powerful tool to strengthen reasoning in this process. Here, we present the AiMS framework, which adapts the classic plan–monitor–evaluate cycle of metacognition to the context of experimental design. This framework emphasizes three iterative stages—Awareness, Analysis, and Adaptation—that scaffold reflection on an experimental system defined by its Models, Methods, and Measurements and evaluated through Specificity, Sensitivity, and Stability. We illustrate application of the AiMS framework through an interactive neuroanatomy case study and provide a structured worksheet to guide readers in applying it to their own experiments. We also highlight how the framework can assist researchers in organizing their ideas for research proposals and explore the responsible use of generative AI as a metacognitive partner that supports structured reflection without supplanting original intellectual contributions. The AiMS framework complements other principles and practices of rigor by foregrounding deliberate reasoning about assumptions, vulnerabilities, and trade-offs. Our goal is to provide practical tools that foster rigor, creativity, and adaptability in the design of biological experiments, supporting both trainees and their mentors in cultivating reflective scientific practices.
Significance Statement
Rigor in experimental design is essential for neuroscience and the life sciences more broadly. The AiMS framework provides a structured approach to reflection, helping researchers to think more deliberately about their design choices, identify underlying assumptions, and assess potential vulnerabilities. By emphasizing metacognition, the AiMS framework promotes experimental rigor by offering practical tools for trainees designing experiments and an accessible scaffold for mentors teaching experimental design.
Introduction
Research design is a fundamental skill for life science trainees (National Academies of Sciences, Engineering, and Medicine, 2018; Verderame et al., 2018; X. Li et al., 2025), yet it is often one of the hardest to master. Designing experiments requires moving beyond executing generic protocols to making deliberate choices about how evidence will be generated and interpreted. As early-career scientists transition from structured coursework to self-directed research inquiries, they face increased responsibility for designing experiments, though oftentimes without having been given clear guidance for making these decisions.
Metacognition—thinking about one's own thinking—offers a powerful lens for scientists to become more aware of how we design, evaluate, and refine experiments. Past work has shown that metacognitive strategies strengthen critical thinking, problem-solving, and decision-making (Tanner, 2012; Zohar and Barzilai, 2013; Stanton et al., 2021). These findings highlight a key point: structured reflection supports better decisions. In the context of experimental design, where uncertainty and multiple possible paths are inherent, metacognition can provide a practical means for scientists to pause, examine our reasoning, and respond more deliberately to unexpected outcomes. Indeed, metacognitive reflection exercises to enhance experimental design have shown promise within course-based undergraduate research experiences (McCabe and Olimpo, 2020).
In this article, we present the AiMS framework, a reconceptualization of a common metacognitive model tailored to the experimental design context. Through this framework, we aim to promote rigorous experimental design through structured reflection. We also highlight how this framework can inform research proposal writing and explore how generative AI, when used responsibly as a metacognitive partner, can support such reflection. We hope this article is of practical use to both trainees and other researchers directly, as well as to science educators and research mentors as a resource for teaching experimental design.
Introducing the AiMS Framework: Structured Metacognitive Reflection to Enhance Experimental Rigor
Here we introduce the AiMS Framework (Fig. 1), through which we have adapted principles from metacognition and experimental design to scaffold thinking and promote structured reflection about experimental systems. At the heart of the framework are the Three A's: Awareness, Analysis, and Adaptation. This metacognitive cycle, adapted from the classic plan–monitor–evaluate model of metacognition that describes activities to regulate one's thinking and learning (Schraw and Moshman, 1995; Stanton et al., 2021), guides researchers to identify key features of an experimental system (“Awareness”), interrogate its limitations and possible experimental outcomes (“Analysis”), and refine the experimental design in light of this reasoning (“Adaptation”).
Conceptual diagram of the AiMS framework. The framework engages scientists in a metacognitive cycle adapted to experimental design. Through the Three A's—Awareness, Analysis, and Adaptation—scientists reflect on aspects of their experimental system, which is conceptualized in this framework as being composed of the Three M's—Models, Methods, and Measurements—that can in turn each be described in terms of the Three S's—Sensitivity, Specificity, and Stability. An editable blank AiMS worksheet is available as Extended Data Figure 1-1 and a completed AiMS worksheet for the neuroanatomy case study is available as Extended Data Figure 1-2.
Figure 1-1
Template AiMS Worksheets. Trainees and educators can download editable worksheets that guide users through the reflective prompts of the AiMS framework. Download Figure 1-1, DOCX file.
Figure 1-2
Completed AiMS Worksheet for Judy’s Neuroanatomy Case Study. This completed worksheet summarizes the neuroanatomy case study presented in the text as a concrete example of application of the AiMS framework to an experimental question and system. An editable version of the worksheet can be downloaded and modified to provide partially worked examples in an instructional context. Download Figure 1-2, DOCX file.
To further structure thinking on experimental systems within this metacognitive cycle, we conceptualize an experimental system—everything that a scientist uses to conduct an experiment (Glass, 2014)—as consisting of the Three M's: Models, Methods, and Measurements. “Models” refer to the biological entities or subjects under study, such as in vitro models (e.g., cell culture, organoids) or in vivo models (e.g., C. elegans, M. musculus). “Methods” are the experimental approaches or perturbations applied to those models—genetic manipulations (e.g., siRNA, CRISPR-Cas9), pharmacological interventions, etc. “Measurements” are the specific readouts or data collected, whether gene expression analyses (e.g., qPCR, RNA sequencing), protein quantification (e.g., Western blot, ELISA), or metabolomics (e.g., mass spectrometry). Thinking in terms of the Three M's makes the moving parts of an experiment explicit, rather than treating the experimental system as a black box.
In turn, each of the Three M's can be described and evaluated through the lens of the Three S's: Specificity, Sensitivity, and Stability. The Three S's capture key elements of established frameworks for experimental design (Glass, 2014), highlighting properties of the system that can shape how results can be interpreted. “Specificity” asks whether the experimental system accurately isolates the phenomenon of interest. “Sensitivity” considers the ability to observe a variable of interest at the amount it exists in the experimental system. “Stability” addresses whether the experimental system remains consistent over time and under various conditions. Reflecting upon these can make visible the assumptions and trade-offs built into experimental design choices.
We note that the Three M's and Three S's are not the only way to frame an experimental system. They represent practical heuristics that are easy to learn and apply. Other approaches—such as frameworks emphasizing validity, reproducibility, or feasibility—can be equally valuable (Masca et al., 2015; Holder and Marino, 2017; Williams, 2018) and could be considered alongside the AiMS framing. While the AiMS framework was designed with wet-lab biological research in mind, we encourage readers to consider how it may be adapted to computational, theoretical, or field-based studies. Rather than prescribing specific methods or dictating a single “correct” approach, we hope to provide a conceptual scaffold for structured reflection that can be flexibly applied across research contexts.
We have developed a series of reflection prompts, organized into an AiMS worksheet (Extended Data Fig. 1-1) to guide users step by step through this framework. The next section of this article is structured as an interactive tutorial to support researchers in gaining hands-on practice applying the AiMS framework to their own experimental contexts. To ground the framework in a concrete neuroscience example, we interleave a neuroanatomy case study [based on Zhang and van den Pol (2016) and summarized in Extended Data Fig. 1-2] that illustrates how each stage of reflection can shape experimental decisions.
These materials were developed for the Community for Rigor (C4R) initiative from the National Institute of Neurological Disorders and Stroke, which is developing freely accessible, online educational modules to improve training in the principles of rigorous research. Through the accompanying online module currently in development by C4R, researchers and educators will be able to further engage with the AiMS framework, worksheets, and neuroanatomy case study presented here through interactive digital content designed to be adapted and incorporated into lab meetings or classroom settings. Further information on the C4R initiative and completed educational modules can be found on the initiative's website at https://www.c4r.io.
Applying the AiMS Framework: An Interactive Tutorial
Meet Judy, a first-year PhD student, who is interested in examining the function of tyrosine hydroxylase (TH)-expressing dopaminergic neurons of the arcuate nucleus of the hypothalamus (ARC)—best known for their role in regulating pituitary prolactin synthesis—in regulating energy homeostasis. Judy first seeks to characterize where in the brain the ARC–TH neurons project. Previous work on ARC–TH cells identified axonal projections to the median eminence (ME), but it is unclear whether these cells project elsewhere in the brain. The paraventricular nucleus of the hypothalamus (PVH), a region near the ARC, is important for energy homeostasis and receives axonal input from other ARC neuronal populations. Thus, Judy's working hypothesis is that the ARC–TH neurons project to the PVH in addition to the ME. She plans to conduct a neuroanatomical tracing experiment: injecting an adeno-associated virus (AAV) expressing a Cre-dependent green fluorescent protein (GFP) into the ARC of transgenic TH-Cre mice and using fluorescent imaging to examine where in the brain the GFP-expressing axons project. Each reflection prompt below is accompanied by Judy's reflections as she develops her experimental design via the AiMS worksheet.
Phase 1: awareness
The first stage of the Three A's, Awareness, is about pausing to take stock of the experimental system. Too often, we rush ahead with familiar techniques without first considering what its key features are, what it can reveal, and whether the system is truly suited to the question at hand. Awareness provides the foundation for more informed choices in later stages of analysis and adaptation.
As a precondition to considering your experimental system, you must first define your research question. A well-defined research question serves as the foundation of experimental design, providing a clear anchor for the reflections that follow. Establishing this focus helps ensure that considerations about experimental design choices, assumptions, and possible outcomes remain tied to a central scientific aim. When formulating research questions, we encourage you to consult existing frameworks, such as patient/population, intervention, comparison, and outcome and feasible, interesting, novel, ethical, and relevant. For a summary of these and other frameworks for defining and refining research questions, see Covvey et al. (2024).
Let Us Reflect: As you prepare to articulate your experimental design, take a moment to identify the specific research question you want to solve.
Judy defined her research question as: To which regions of the brain do ARC–TH neurons project?
Step 1 of AiMS worksheet: identity components of your experimental system
The first step to applying the AiMS framework is to identify core components of the experimental system. Explicitly outlining the Three M's—Models, Methods, and Measurements—at play brings into view elements that will shape both what you can learn with your experimental system and how you will interpret your results. Articulating these components provides a foundation for more structured reflection in later steps.
Let Us Reflect: To address your research question, what Models, Methods, and Measurements do you plan to, or could you, use?
Judy identified the following components of her experimental system:
Model: Transgenic TH-Cre mouse
Method: Cre-dependent AAV-GFP viral injection into the ARC of TH-Cre mice
Measurement: Detection of GFP+ fibers in brain sections
Step 2 of AiMS worksheet: examine features of your experimental system
After defining the basic components of the experimental system, the next step is to consider their functional properties. Examining Specificity, Sensitivity, and Stability highlights our underlying assumption and reveals where the system may be particularly robust or vulnerable. This exercise is not intended to yield precise quantification but to surface practical parameters that may influence how we obtain and interpret evidence.
Let Us Reflect: Select one Model, Method, and Measurement that you articulated in Step 1. For each component, describe experimental parameters that map onto Specificity, Sensitivity, and Stability.
Judy's reflections about her experimental system are given in Table 1.
Example reflections for neuroanatomy case study, Step 2 of AiMS framework
The Awareness phase of the AiMS framework emphasizes making deliberate choices in selecting an experimental system. Explicitly identifying its components and considering their key properties can help establish a clearer foundation for design choices and avoid rushing ahead with unexamined assumptions. This groundwork provides the opportunity to ensure alignment of the general approach to the central research question and paves the way for deeper analysis in the next phases of the AiMS cycle.
Phase 2: analysis
With an awareness of the key features of an experimental system, the next step is Analysis. This phase focuses on probing deeper: envisioning possible outcomes, considering how those outcomes might be interpreted, and recognizing where the system might fail. Thus, the Analysis phase helps us think critically about what the experimental system can and cannot tell us.
Step 3 of AiMS worksheet: consider multiple possible experimental outcomes
A useful starting point for analysis is to imagine the full range of possible outcomes an experiment could produce. Rather than focusing only on the anticipated or desired result, it is useful to consider what the experiment would look like if the hypothesis is not supported or if something unexpected occurs. This reflection exercise helps guard against common cognitive biases, such as confirmation bias—the tendency to value results that align with a preferred hypothesis while discounting contradictory evidence (Born, 2024). Shifting away from expecting a single “correct” result productively reframes every outcome, including null or unexpected findings, as potentially informative.
Let Us Reflect: For at least one of your Measurements, describe multiple possible experimental outcomes you might observe. You do not yet need to describe how you might interpret each outcome.
For her neuroanatomical tracing experiment, Judy has identified multiple possible outcomes for her measurement, detection of GFP+ fibers in brain sections:
Possible observation #1: Presence of GFP+ fibers in the ME only
Possible observation #2: Presence of GFP+ fibers in the ME and the PVH
Possible observation #3: Presence of GFP+ fibers in the ME and PVH, as well as additional brain regions
Possible observation #4: Presence of GFP+ fibers in the ME and additional brain regions, but not the PVH
Step 4 of AiMS worksheet: distinguish biological interpretation versus technical artifact
A given experimental outcome is rarely self-explanatory. The same observed outcome could arise from valid biology or from technical artifacts/limitations of the system. A strong signal in the expected direction, for instance, might genuinely reflect the underlying biological process, but it could also arise from off-target effects, cross-reactivity, or unintended variation in the system. Similarly, the absence of a signal might indicate that the hypothesized effect does not exist, or it may simply reflect insufficient sensitivity, inadequate sample quality, or measurement error. By explicitly considering both biological explanation and technical artifact interpretation for any results, we can strengthen our ability to design appropriate controls, interpret data cautiously, and avoid premature or overconfident conclusions.
Let Us Reflect: For each of the possible observed outcomes that you identified, describe at least one biological interpretation of the observation and at least one technical artifact interpretation.
Upon reflection, Judy realized that the possible experimental outcomes she listed might have different interpretations. Her reflections capturing potential biological interpretations versus technical artifact interpretations are given in Table 2.
Example reflections for neuroanatomy case study, Step 4 of AiMS framework
Step 5 of AiMS worksheet: identify potential experimental failure points
Now that we have explored ways that technical artifacts might confound our interpretation of experimental outcomes, let us explore further how the experimental system itself can break down. Weaknesses can arise at the level of the Model, such as when it does not fully capture the phenomenon of interest; at the level of the Method, such as when the intervention is applied inconsistently or with limited efficiency; or at the level of the Measurement, such as when experimental readouts are distorted by noise or signal loss over time. Thinking through these weak points does not assume that experimental failure is inevitable, but it highlights where evidence could be compromised. By acknowledging these risks upfront, we can build appropriate data collection into our experimental design to monitor our experimental system, as discussed in the next section, rather than discovering the problems only after data collection, thereby strengthening the reliability of our results.
Let Us Reflect: Identify ways in which the Model, Method, and Measurement components of your experimental system could potentially fail along the dimensions of Specificity, Sensitivity, or Stability.
Judy noted that her system has certain vulnerabilities. Some of her ideas for how her system could fail are depicted in Table 3.
Example reflections for neuroanatomy case study, Step 5 of AiMS framework
Taken together, these reflections highlight that the Analysis phase is not about predicting a single correct answer. It is about mapping the terrain of possible outcomes, alternative explanations, and failure points. This process is not intended to eliminate uncertainty but make it visible and manageable, paving the way for the final phase of our metacognitive cycle: Adaptation.
Phase 3: adaptation
Having identified key features of an experimental system (Awareness) and anticipated its possible outcomes and failure modes (Analysis), we now transition to the final stage of the Three A's: Adaptation. Adaptation is about acting on reflection: introducing strategies that strengthen interpretation, recognizing when technical limitations may be insurmountable, and making deliberate choices about how to proceed.
Step 6 of AiMS worksheet: plan to monitor performance of your experimental system
As discussed previously, a recurring challenge in experimental design is distinguishing whether a result reflects true biology or technical artifact. We can reduce this uncertainty by “collecting data”—through the design of experimental controls—on how the experimental system itself behaves during the experiment. “Collecting data” to probe the Specificity, Sensitivity, and Stability of your experimental system—such as by validating reagents, determining dynamic range, or assessing replicability across contexts—can further clarify how well the experimental system is performing. These steps do not eliminate uncertainty, but they increase confidence that results are interpretable in biological terms rather than reflecting system artifacts.
Let Us Reflect: Select a subset of potential technical failure modes you identified during Step 5. Brainstorm ways you could “collect data” on whether the technical failure might have occurred when you run your experiment.
Judy reflected on how to monitor the performance of her experimental system. Her description of some of the potential technical failures she might encounter and ways to “collect data” to monitor whether these failures might have occurred are summarized in Table 4.
Example reflections for neuroanatomy case study, Step 6 of AiMS framework
We have structured our discussion here to encourage broad thinking about how you could monitor the performance of your experimental system without imposing constraints of precise labeling or classification of specific experimental control conditions. However, such classification can be helpful as you continue to build out your experimental approach. We encourage you to explore some of the many excellent educational resources that elaborate upon different types of experimental controls (Glass, 2014; Harrington, 2020; Community for Rigor, n.d.) and to incorporate relevant concepts into your fully articulated experimental plan.
Step 7 of AiMS worksheet: identify potential experimental pitfalls and alternative experimental approaches
All experimental systems have their technical limitations, and some technical limitations cannot be addressed or overcome through experimental controls that provide additional data on the experimental system. These potential experimental pitfalls represent possibly insurmountable technical constraints that would prevent a system from addressing the research question. Reflecting upon potential pitfalls can allow us to avoid investing resources in a design that cannot yield interpretable answers. In some cases, this may mean redefining the research question; in others, it may involve a shift to an alternative experimental approach with a different Model, Method, or Measurement that enables you to bypass technical limitations of your initial experimental system.
Let Us Reflect: Which of the technical limitations that you identified for your experimental system are most likely to pose technical pitfalls that are not easy to overcome through troubleshooting? Also consider other aspects of your experimental system not captured in the previous reflection exercises that may also present potential experimental pitfalls (e.g., inherently challenging experiments to execute, untested approaches). What alternative approaches could enable you to overcome the potential experimental pitfalls that you identified? Try to identify at least one alternative model, method, and measurement.
Judy recognized several potential experimental pitfalls of her system and began considering alternative experimental approaches to answer her research question. A sampling of her ideas is presented in Table 5.
Example reflections for neuroanatomy case study, Step 7 of AiMS framework
Step 8 of AiMS worksheet: apply a critical lens to your initial and alternative experimental approaches
Notably, exploring alternative approaches does not always mean replacing one approach with another. Complementary strategies can strengthen findings by corroborating results across methods or adding new layers of evidence. In neuroscience research, for example, optogenetic and chemogenetic approaches are often paired (Krut’ et al., 2024; Meron Asher and Goshen, 2025; S. Li et al., 2025); viral neuroanatomical tracing can be combined with electrophysiology (Kim et al., 2017), and calcium imaging is interpreted alongside electrophysiology recordings (Wei et al., 2020). Emerging technologies like spatial transcriptomics illustrate how integrating distinct methods can leverage the relative advantages of both methods, underscoring the value of complementary techniques (Williams et al., 2022). Considering such complementary approaches at the design stage allows us to plan for both corroboration and elaboration, increasing confidence in our conclusions.
At the same time, reflecting on alternative experimental approaches provides an opportunity to critically examine the factors that drove our initial selection of the experimental system. Such factors can include both scientific rationale (e.g., perhaps your method offers the greatest sensitivity) and legitimate logistical reasons (e.g., perhaps the mouse model you are using was already available at your institution), but it might also be influenced by cognitive biases (Do you perceive a particular experimental method or analysis approach to be “superior” just because it has been extensively used in your laboratory? Have you taken the time to reflect on whether it is actually the best-suited approach?). Thus, we provide a final prompt to structure reflection on your initial experimental system to further mitigate the influence of cognitive biases and increase the rigor of your approach.
Let Us Reflect: What were the primary factors driving your selection of your initial experimental Models, Methods, and Measurements? What factors led you to not pursue your proposed “alternative” approaches in your initial experimental design? Are there any advantages to trying your “alternatives” from the outset or in parallel to your initial experimental design as a complementary approach?
Judy generated many insights about her initial experimental design decisions, some of which are presented below:
She deemed the specific viral tracing strategy using a Cre-dependent virus injected into the ARC of TH-Cre mice to be the best approach to selectively express the GFP tracer in the dopaminergic cells of the ARC to characterize their axonal projections in a specific, yet unbiased, fashion. If she had crossed the TH-Cre mouse to a Cre-dependent GFP reporter line, she would have lost the spatial specificity and would have indiscriminately labeled all TH+ neurons, including those in other parts of the hypothalamus. She selected the specific TH-Cre transgenic mouse line based on its availability in a commercial mouse repository and because it had been previously characterized in published research. She initially chose to measure the presence of GFP+ fibers in brain sections because it seemed sufficient to answer her research question; a large body of prior neuroanatomical studies had characterized axonal projection patterns via fiber detection in brain slices.
Judy did not initially pursue whole-mount brain imaging to characterize the pattern of ARC–TH axonal projections not only because of the precedent that imaging of brain sections would be sufficient but also because it would require the development of novel brain clearing and imaging protocols in her lab and access to a light sheet microscope, which was not readily available in her department.
Judy described using a retroAAV viral anatomical tracing approach—in which she would inject a Cre-dependent retroAAV at the predicted sites of ARC–TH axonal terminals to retrogradely label their cell bodies—as a potential alternative approach to her anterograde tracing strategy of injecting ARC–TH neuron cell bodies to label their axons. As a primary strategy, this approach has the limitation of being candidate-based, as opposed to unbiased, only returning information about whether ARC–TH neurons project to the specific regions where the virus was introduced. However, this would be a useful approach to validate potential target regions identified from the anterograde tracing experiments or could even be used in parallel with the anterograde tracing strategy to test Judy's specific hypothesis that ARC–TH neurons project to the PVH.
Thus, we complete a full revolution of our AiMS metacognitive cycle. Through the Adaptation phase, you make explicit decisions based on what has been learned through Awareness and Analysis while also identifying alternative experimental approaches. This, in turn, should trigger a return to the Awareness and Analysis phases for further reflection to refine your experimental design before such alternative experimental approaches are implemented. Importantly, the iterative nature of this metacognitive process is meant to guide, not paralyze, you. Experimental design often involves committing to a design despite known limitations, provided those limitations are acknowledged and managed. It may also involve pivoting to alternative strategies or layering complementary approaches to strengthen conclusions. What matters is not that every design is perfect—few, if any, ever are—but that choices are made deliberately, with a clear-eyed view of the assumptions, vulnerabilities, and trade-offs involved.
From Structured Thinking to Structured Writing: Application to Research Proposals
Fellowship and grant proposals provide one of the primary venues where experimental design is formally articulated. These documents emphasize brevity, typically imposing strict word or page limits that preclude a comprehensive description of every experimental design choice, assumption, and possible alternative. Nevertheless, completing the type of structured reflection outlined in the AiMS framework remains valuable. Even if the full details of your responses cannot be included in a written research proposal, these reflections can help clarify the logic of the experiment and make it easier for you to distill the most critical points for reviewers.
In our experience teaching graduate-level research proposal writing courses, we have observed recurring misconceptions that the AiMS framework is designed to counteract. An often-encountered example is framing experiments as binary verdicts—“If I observe [the predicted outcome that aligns with my hypothesis], I will conclude [the biological interpretation]. Otherwise, I will conclude that my experiment has failed.” Closely related, students sometimes define “pitfalls” as the absence of their predicted outcome—“a potential pitfall is that I won't observe the outcome that is aligned to my hypothesis.” This flawed reasoning overlooks the possibility that the same observation could reflect either an underlying biological reality or a technical artifact of the system and that unexpected or null results can still be informative if properly controlled for and interpreted. The AiMS framework—through guided reflections on anticipating multiple outcomes, distinguishing biological interpretation from technical artifacts, and identifying genuine pitfalls—was designed to mitigate these misconceptions, thereby supporting more rigorous reasoning in research proposals.
Indeed, information from the completed AiMS worksheet maps readily onto the structure of research proposals. Proposal formats commonly require a clear statement of the research question, description of the experimental system, discussion of experimental controls, identification of multiple possible experimental outcomes and their interpretations, and consideration of potential pitfalls and alternative experimental approaches. Each of these aligns directly with reflection prompts in our AiMS framework. Thus, engaging with the worksheet before drafting a proposal can help streamline the writing process: the reflective work provides a reservoir of ideas and reasoning that can be selectively condensed into the research proposal format while ensuring that the proposal conveys to reviewers that the experiment has been carefully thought through in terms of both strengths and possible limitations. As we describe in the following section, generative AI tools can also play a role in this process, serving as a metacognitive partner that helps organize reflections from the AiMS framework into proposal-ready formats while preserving the original contribution of the researcher.
Generative AI as a Metacognitive Partner
Generative AI (Gen AI) tools, such as ChatGPT, present both opportunities and challenges as they become integrated into the scientific discovery process (Wang et al., 2023; Kwon, 2025; Naddaf, 2025). The current Gen AI landscape is evolving quickly. Multiple funding agencies, publishers, and institutions have recently issued guidance on the appropriate use of Gen AI, emphasizing transparency and disclosure. For example, the National Institutes of Health (NIH) clarified in a July 2025 Notice (National Institutes of Health, 2025) that grant proposals must reflect the original intellectual contribution of the applicant(s) and should not be “substantially developed by AI.” Multiple journals, including eNeuro and its parent Society for Neuroscience, have adopted the International Committee of Medical Journal Editors recommendation that AI-assisted writing must be disclosed (ICMJE, n.d.). Ethical concerns further complicate the use of Gen AI tools, including known biases inherent in large language models (LLMs), legal questions of copyright and ownership, risks to data privacy when sharing sensitive or unpublished materials, and the significant environmental footprint of these tools (Weidinger et al., 2021; Masters, 2023; Tran et al., 2025). Taken together, these developments reflect a new normal in scientific research: Gen AI tools should be used to “enhance,” but not “replace,” the original contributions of researchers and must be engaged responsibly and transparently.
Within this context, both our explorations and recent work demonstrating the strong metacognitive capabilities of Gen AI LLMs (Pavlovic et al., 2024) suggest that Gen AI holds promise for enhancing the rigor of experimental design when used as a “metacognitive partner,” not by producing answers but by encouraging deeper and more structured reflection. For example, when we prompted ChatGPT with “I am going to describe my experimental design for a neuroscience experiment. Act as a metacognitive partner who prompts me with questions to probe aspects of my design to enhance the experimental rigor and combat possible cognitive biases embedded within the design,” its output included thoughtful questions spanning multiple aspects of the experimental design (see Fig. 2 for an excerpt of the interaction with ChatGPT). This interaction demonstrates how Gen AI can be recruited to strengthen reflective practices and make the experimental design process more deliberate.
A demonstration of using generative AI as a metacognitive partner in experimental design. Screenshots depicting excerpts from an interaction with ChatGPT5 (university enterprise account) illustrate how generative AI can facilitate metacognitive reflection. Note that ChatGPT responses are stochastic and context-dependent, so this output is a representative example that is unlikely to be reproduced verbatim by other users.
Building on this example, we found that Gen AI can take on several roles that enhance experimental design using the AiMS metacognitive framework: as a brainstorming partner, a constructive reviewer, or an organizing assistant. Each of these roles is described in general terms below; specific prompts for use with Gen AI that reflect each of these roles, as well as prompts to critically engage with your Gen AI partner, are given in Box 1. As a brainstorming partner, Gen AI can scaffold the reflection process through structured prompts that surface assumptions and clarify reasoning. You can also prompt it to expand on initial ideas, propose alternatives, and generate suggestions under specific constraints. In this role, Gen AI does not replace original thinking but can help deepen and broaden reflections, encouraging considerations beyond your first instincts. As a constructive reviewer, Gen AI can be asked to critique an experimental plan from the perspective of a grant reviewer, journal referee, or a general critic. It can point to missing controls, highlight how confirmation bias might influence interpretation, or flag other potential weaknesses. In some cases, it can help prioritize which vulnerabilities are most critical to address. It can even be prompted to critique its own outputs, a meta-exercise that can improve the quality or breadth of the responses. Used this way, Gen AI functions not as an idea generator but as a critic that encourages you to evaluate your design from the standpoint of rigorous peer review. As an organizing assistant, Gen AI can help map content from a completed AiMS worksheet into other formats, such as experimental proposals or manuscript methods sections. This may include aligning your reflective responses with reporting guidelines such as the ARRIVE guidelines (Percie du Sert et al., 2020) or the Cell Press STAR Methods (Marcus, 2016). By organizing and reformatting information, GAI can make the underlying logic of your experimental design clearer to others while still preserving your original intellectual contribution.
Example prompts for using generative AI in experimental design. Generative AI can adopt different roles and support experimental design by facilitating metacognitive reflections. Below are example prompts that align with the roles of a brainstorming partner, a constructive reviewer, and an organizing assistant, as well as ones that prompt Gen AI to appraise its own output. These prompts have been tested with a combination of ChatGPT models 4o and 5. These prompts can be adapted to individual research contexts or serve as inspiration for new prompts.
The brainstorming partner
Goals: broaden ideas and surface overlooked considerations
“I am going to describe my experimental design for a neuroscience experiment. Act as a metacognitive partner who prompts me with questions to probe aspects of my design to enhance the experimental rigor and combat possible cognitive biases embedded within the design.”
“Using structured prompts and questions, guide me to complete the attached worksheet for my specific experimental system. You should provide guiding questions but do not actually complete the table. This is meant to be an exercise to guide my thinking, but not to provide me with the answers.” [Upload blank AiMS worksheet]
“Review the attached experimental plan. Using your expertise as a neuroscience researcher, elaborate on each component of this worksheet and regenerate an expanded version with additional examples provided for each prompt.” [Upload completed AiMS worksheet].
“For my model, method, and measurement, can you identify additional examples of specificity, sensitivity, and stability?” [Upload completed table for reflection prompt #2.]
“I am a researcher who wants to investigate a similar research question as that described above (where in the rodent brain do ARC–TH neurons project) but my lab is resource-limited and we don't have the biosafety clearance to perform viral neuroanatomical tracing experiments. Propose equally rigorous experimental approaches that I could use instead.”
“I am a graduate researcher writing my first research proposal. What are some common misconceptions held, or errors made, by research trainees related to (1) identifying multiple possible outcomes to their experiments, (2) identifying potential pitfalls to their experiments, and (3) proposing alternative approaches when writing their research proposals? Propose strategies that I can take to avoid making these common mistakes.”
The constructive reviewer
Goals: probe weakness or biases and provide constructive feedback
“You are a leading neuroscience researcher and expert in principles of experimental rigor. Review the attached experimental plan and generate a series of guiding questions to be delivered as feedback to the researcher to highlight areas that need to be strengthened.” [Upload completed AiMS worksheet]
(To prompt GAI to evaluate its own output.) “You are now a reviewer analyzing an experimental design that includes the experimental controls that you just suggested. Provide constructive feedback as to what is missing and what technical artifacts may still arise, despite the inclusion of these controls. Provide concrete suggestions for how the experimental plan could be strengthened according to published frameworks of rigorous experimental design. Provide real and accurate citations.”
“Review my uploaded experimental plan. How might my brainstorming about possible experimental outcomes be influenced by confirmation bias or other cognitive biases?” [Upload completed AiMS worksheet]
“Analyze the potential technical limitations or failure modes described in the attached experimental plan and prioritize them based on the likelihood of their occurrence. What aspects of experimental design should I prioritize to address the most likely technical issues?” [Upload completed AiMS worksheet]
“You are a professor chairing a preliminary qualifying examination committee for a PhD student. Use the uploaded AiMS worksheet to generate probing questions about the experimental design to test the PhD student's knowledge of their experiment and principles of experimental rigor. Give one question at a time and then respond to my answer as if we are engaged in the questioning during the examination.” [Upload completed AiMS worksheet]
The organizing assistant
Goals: organize ideas and reframe information
“You are a professor chairing a preliminary qualifying examination committee for a PhD student. Use the uploaded AiMS worksheet to generate probing questions about the experimental design to test the PhD student's knowledge of their experiment and principles of experimental rigor. Give one question at a time and then respond to my answer as if we are engaged in the questioning during the examination.” [Upload completed AiMS worksheet]
“I'm attaching a worksheet that describes aspects of my experimental design. Map this information onto an outline for an NRSA-style NIH research proposal, noting which information I should prioritize for inclusion in the proposal given the restricted page limit and how I should organize it according to standard grant headings.” [Upload completed AiMS worksheet]
○ Alternative: “Re-do the prior analysis. Rather than directly mapping the content of the worksheet onto an NRSA-style NIH proposal outline, generate a series of guiding questions that prompt me to do so.”
“Please map the information contained within the attached worksheet onto the ARRIVE guidelines (described here: https://arriveguidelines.org/). Highlight gaps in my experimental plan that I need to address to meet the ARRIVE guidelines for reporting.” [Upload completed AiMS worksheet]
Critically engaging with Gen AI
Goals: prompt Gen AI to evaluate its own output and provide tools for human verification of AI
“Perform an honest and self-critical appraisal: what are the relevant strengths and limitations of generative AI tools like ChatGPT 5 to assist human scientists in experimental design?”
“I'm using genAI to help create metacognitive worksheets to guide scientific trainees in experimental design. Identify the ways in which genAI might NOT be helpful toward this goal.”
○ Follow-up: “For each of the above limitations given, provide some metacognitive reflection questions that engage the learner in thinking to mitigate the limitation.”
“Please attempt to validate each of the citations you provided above. If incorrect, please identify the correct citations.”
“Generate a checklist for scientific trainees for how to vet the information that is generated from GenAI.”
Despite these potential benefits of using Gen AI in supporting experimental design, it also poses important limitations to keep in mind. In a meta-reflective exercise, we asked ChatGPT5 to perform a self-critical appraisal of its strengths and limitations in assisting researchers in experimental design. Its output (Fig. 3) echoes our own observations and aligns with known limitations of Gen AI, such as producing generic responses—which could result in an overall loss of creativity—plausible but inaccurate information and hallucinated citations (Weidinger et al., 2021; Huang et al., 2025; Peters and Chin-Yee, 2025). Ultimately, researchers are responsible for the accuracy and integrity of any information that Gen AI may supply. For this reason, careful verification, critical evaluation, and alignment with institutional and disciplinary policies are essential. Responsible use requires informed decisions that reflect both professional standards and values.
Strengths and limitations of using gen AI in experimental design, according to ChatGPT. A screenshot of the table generated by ChatGPT5 (university enterprise account) summarizes the output in response to the prompt: “Perform an honest and self-critical appraisal: what are the relevant strengths and limitations of generative AI tools like ChatGPT 5 to assist human scientists in experimental design?”.
Our experience exploring the use of Gen AI in experimental design, consistent with others’ (Steinauer, 2025), suggests that Gen AI is most effective when used as a sparring partner rather than as an answer generator. Although the speed of Gen AI may tempt you to seek quick solutions, engaging with it might be most valuable when it slows you down, prompting you to question assumptions, examine alternatives, and strengthen your experimental design. In this way, using Gen AI as a metacognitive partner reinforces the central theme of the AiMS framework: structured reflection, not shortcuts, is what ultimately strengthens experimental design.
Conclusion
This paper introduces the AiMS framework as a structured approach to experimental design rooted in metacognitive reflection. The Three A's—Awareness, Analysis, and Adaptation—provide a reflective cycle for recognizing the features of an experimental system, envisioning outcomes and limitations, and refining strategies in response. The Three M's—Models, Methods, and Measurements—help disentangle the moving parts of an experimental system, while the Three S's—Specificity, Sensitivity, and Stability—offer heuristics for evaluating their strengths and weaknesses. Together, these tools encourage researchers to move beyond procedural execution toward deliberate, reflective choices in experimental design that acknowledge vulnerabilities and trade-offs. We note that the AiMS framework does not capture many essential principles of rigorous experimentation, such as randomization and blinding, the use of replicates, and best practices for data analysis. These principles and practices, described in detail in many excellent resources (Glass, 2014; Harrington, 2020; Community for Rigor, n.d.; National Institute of General Medical Sciences, n.d.), remain indispensable. The AiMS framework is intended to complement, rather than replace, such standards by foregrounding structured reflection on experimental systems. While demonstrated here with a neuroanatomical case study, the AiMS framework is adaptable to other areas of neuroscience and other domains of life science research. We also explored how generative AI, when used responsibly as a metacognitive partner, can extend this reflective process without supplanting original intellectual contribution. Taken together, our goal is not to prescribe rigid rules but to provide tools that help researchers think more critically and creatively about their experimental design choices. Looking ahead, integrating structured reflection practices alongside established best practices and engaging responsibly with emerging technology can help foster rigor, creativity, and adaptability in the evolving landscape of scientific discovery.
Generative AI Usage Statement
Generative AI (ChatGPT4o and 5, OpenAI) was used in testing the application of the AiMS framework and examining its potential to serve as a metacognitive partner in experimental design. Some of its outputs are included in the text as examples and are clearly indicated. During the manuscript writing process, ChatGPT5 was occasionally used for text editing. All content of this manuscript has been critically reviewed, revised, and approved by the authors, who take full responsibility for the accuracy and integrity of the manuscript.
Footnotes
The authors declare no competing financial interests.
This article was based on educational materials developed by the authors for the “Center for Rigor” (C4R) consortium funded by METER award 5UE5NS128294 from NIH/NINDS. We thank our colleagues from our HMS METER team, Rick Born and Ella Batty, and Kevin Rusch from C4R for early brainstorming and feedback on the materials. We also thank Deepali Ravel, Alan Emanuel, and Ian McLachlan for providing helpful comments on the manuscript.
This is an open-access article distributed under the terms of the Creative Commons Attribution 4.0 International license, which permits unrestricted use, distribution and reproduction in any medium provided that the original work is properly attributed.
References
In this issue
Jump to section
- Article
- Abstract
- Significance Statement
- Introduction
- Introducing the AiMS Framework: Structured Metacognitive Reflection to Enhance Experimental Rigor
- Applying the AiMS Framework: An Interactive Tutorial
- From Structured Thinking to Structured Writing: Application to Research Proposals
- Generative AI as a Metacognitive Partner
- Conclusion
- Generative AI Usage Statement
- Footnotes
- References
- Synthesis
- Author Response
- Figures & Data
- Info & Metrics
- eLetters