Best Practices

Best Practices for Successful Multidisciplinary Collaborations

The U.S. Department of Energy’s Scientific Discovery through Advanced Computing (SciDAC) program brings together many of the nation’s top researchers in applied mathematics, computer scientists, and domain scientists to collaborate and solve some of the most challenging scientific problems using advanced computing capabilities that are available at the Department of Energy (DOE). Such collaborations involve teams, each of which is composed of researchers with very different expertise. The multidisciplinary nature can become difficult if not managed well. Here, we have collected some best practices from experienced members of the SciDAC community that, if followed, hopefully will lead to productive and successful multidisciplinary collaborations.

The materials below are drawn from the discussions in a panel, “Best Practices for Successful Multidisciplinary Collaborations”, held at a SciDAC Principal Investigators Meeting on September 18, 2025. The panelists were long-time participants in the SciDAC program with diverse backgrounds and were involved in many multidisciplinary collaborations in SciDAC.

SciDAC teams are interdisciplinary, across applied mathematics (AM), computer science (CS), and domain science (DS) topics that are central to the mission of DOE programs and offices (e.g., BER, BES, FES, HENP, NE and OE). Teams are comprised of CS/AM experts that usually belong to the ASCR SciDAC institutes and DS investigators. Within the SciDAC program, the discovery process in domain science is enabled by the synergies between CS/AM experts and DS investigators.

Q: Why are multidisciplinary collaborations important in SciDAC and what are the intellectual benefits?

• Solving Complex Problems: By bringing together diverse expertise (e.g., domain science with applied mathematics and computer science, as well as advanced computing), problems that are intractable for a single discipline can be addressed.
• Maximizing Research Impact: By ensuring computational tools are well-aligned with scientific questions and vice-versa, the research outcomes become more relevant and powerful.
• Driving Innovation: The cross-pollination of ideas and methods between different fields leads to novel approaches and discoveries that wouldn't emerge in siloed research.
• Creating New Tools and Methods: The discussions around “what tools do you wish existed“ and the need for new approaches (e.g., in data challenges or even building common language with artificial intelligence (AI)) point to the intellectual benefit of driving tool and methodology development.
• Workforce Training: Developing and training individuals to effectively contribute to interdisciplinary computational research by interacting with both CS/AM researchers and DS investigators.

Q: What should effective multidisciplinary collaborations look like?

• Built on Strong Communication: This is considered the “heart“ of a multidisciplinary collaboration. It involves investing time in mutual education, creating an open environment for questions, and consciously working to bridge linguistic and conceptual gaps between fields.
• Patience and Accepting of Longer Timelines: Effective interdisciplinary collaborations take time to “click“ (6 months to 2 years). Multidisciplinary work inherently has a slower pace, including non-publishable efforts.
• Characterized by Mutual Understanding and Respect: Both CS/AM and DS researchers should strive to deeply understand each other's problems, constraints, and “languages.“
• Strategically Managed with Clear Goals: Investigators should identify common scientific “pain points“ and set clear near-, mid-, and long-term goals. Focus should be maintained on key deliverables, especially when challenges arise.
• Supported by In-Person Interaction: Regular in-person meetings, especially at the start, are vital for building rapport, trust, and ensuring alignment, complementing online communication.
• Flexible and Adaptable: Teams must be willing to change course, adapt to new challenges, and evolve their approach as the project progresses.
• Generous with Recognition: Fair co-authorship and understanding the importance of different publication venues across fields contribute to a healthy, effective collaboration.

Q: What are the biggest challenges in communicating across different scientific “languages” (e.g., physics vs. computer science)?

• Time Investment: Getting people to dedicate enough time to truly understand each other's domains is difficult because individuals are often spread across many projects.
• Initial Education Phase: Accept that time needs to be spent, particularly at the beginning of the project, just educating both sides on terminology, notation, and the inherent difficulties of their respective fields.
• Differing Perceptions of "Hard" vs. "Easy": What one discipline considers “hard” or “easy” may be completely different for another, leading to misunderstandings and missed opportunities.
• Overloaded Terminology: Technical but also common words can carry different meanings in different fields, causing miscommunication.
• Translating Problems: Being able to present problems from different domains in a way that is comprehensible to others can be very difficult, stressing the need for individuals who can “translate.”

Q: How do you ensure that CS/AM researchers understand the scientific questions deeply enough, and DS investigators understand the mathematical and computational constraints?

• Dedicated Time for Mutual Education: Investing time and energy at the beginning is critical for education on both sides, ensuring they learn each other's terms, notation, and difficulties.
• Finding Common Ground: Finding a common language or background to facilitate understanding is critical. For example, having a CS person with physics background may help.
• Open Communication and Asking Questions: Fostering an environment where everyone feels comfortable asking questions, which directly helps clarify scientific questions for CS/AM experts and computational constraints for DS investigators. Keep in mind that something sounding simple and straightforward to CS/AM researchers may look difficult to domain scientists, and vice versa.
• "Translators": It is useful to have individuals in a collaboration who possess enough knowledge in both areas to effectively bridge the gap and translate between the domains, ensuring deeper understanding.
• Tutorials and Demonstrations: Some partnerships use sets of tutorials showing science code and demonstrations to showcase capabilities. These methods provide concrete ways for AM/CS researchers to grasp scientific context and for DS investigators to understand computational capabilities and limitations.

Q: What role does training or cross-disciplinary education play in bridging these gaps? How do you keep the process efficient?

• Direct Role in Bridging Gaps: Training and cross-disciplinary education play a crucial and direct role.
• Tutorials: Tutorials to allow people to ask basic questions and build foundational understanding.
• HPC Training for Domain Scientists: Early-career domain scientists to attend HPC trainings, directly enabling them to better understand computational aspects.
• Seeking Dual Expertise: The value of cross-disciplinary education by advocating for people who know enough about both sides to translate is crucial.
• Keeping the Process Efficient so that team members won’t find it time consuming.
• Time Investment (Paradoxically): While it seems counterintuitive for efficiency, dedicating sufficient initial time to cross-disciplinary education and communication prevents inefficiencies later on due to misunderstandings, rework, or stalled progress.
• Avoiding "Paper Dumps": Efficiency comes from targeted educational efforts like tutorials, rather than simply bombarding team members with foundational papers.
• Focused Communication: By dedicating time to mutual understanding, the team avoids false starts and ensures that efforts are aligned from the outset.
• Clear Goals and Deliverables: Keeping eyes on key deliverables and focusing on the bare minimum for success also contributes to efficiency by preventing scope creep or getting bogged down in non-essential tasks.

Q: What collaboration models (embedded teams, partnerships, co-located research groups) have worked best?

• In-person interaction (especially for starting projects): In-person interaction is an amazing way for a good start of a project. The physical presence can avoid assumptions about behaviors and fosters rapport and trust. Both are critical for an effective collaboration.
• Regular online meetings: Regular online meetings are important as well, particularly for teams with members residing in different locations. While in-person is crucial for foundational elements, consistent virtual interaction helps maintain momentum and coordination.
• "Embedded Teams": The “Embedded Teams” model is a great way to involve members from different disciplines working closely within a single team structure.
• Visits for multiple weeks: Having people visit for multiple weeks provides a way for a deeper, more sustained in-person engagement than just a kick-off meeting.
• Showing capabilities through demonstrations: Going through science codes, tutorials and demonstrations to show capabilities can be very valuable. It provides a powerful way to bridge understanding and foster collaboration, regardless of the precise team structure.

Q: How do you overcome challenges?

• Mutual Education: Establishing an understanding of terms and notation in a multidisciplinary collaboration to bridge knowledge gaps is fundamental to a productive and successful interaction.
• Flexibility and Adaptability: Be flexible when facing roadblocks or shifting goals. Recognize when things need to change or goals need to shift. Be able to accept change or shift.
• Focus on Key Deliverables: Stay focused on key deliverables when things got bogged down. Sometimes focusing on bare minimum may help achieving success, which in turn may maintain momentum.
• Strategic Stepping Back: Another possibility of overcoming roadblocks is to take a step back and view where things are and where things are going. Ask what steps are needed for the next success. This helps re-evaluate and refocus the team.
• Dealing with Bottlenecks: When difficulties or bottlenecks arise, a useful way to help resolve conflicts and maintain progress is by finding common ground.
• Prioritizing Domain Science Needs: Prioritizing the discovery goals can be a way to overcome misalignments. While meeting DS investigators’ requests is important, the balance with foundational work in CS/AM should be considered to advance techniques and/or algorithms.
• Acknowledging Foundational Work: It is not uncommon for high-impact, cutting-edge research to begin with preliminary work that may feel boring. Managing expectations and validating the necessity of less glamorous tasks can be useful.
• Understanding Time Horizons: A multidisciplinary collaboration typically proceeds with a slower timeline because of, for example, the overhead in establishing a common language and mutual understanding of the problem to be tackled. Setting realistic expectations about project duration and when results can be expected would be crucial to avoid misalignment.
• Establishing Near/Mid/Long-Term Goals: CClear goal setting from the outset helps align expectations. That includes establishing near-, mid-, and long-term goals.
• Getting Actively Involved: A proactive approach to understanding and resolving issues is through direct engagement.
• Regular In-Person Interactions (for remote work): It is generally easy to lose momentum and cohesion in a collaboration that involves people from different locations. Regular in-person interactions provide a way to maintain momentum and cohesion.

Q: How long has it taken for the collaborations to click and become productive?

• It is not unusual for a new multidisciplinary collaboration to take 6 months to 2 years to become truly productive. Existing teams may potentially integrate new members faster.

Q: What are the strategies/best practices for productively engaging in multidisciplinary collaborations?

The panelists offered several direct and indirect strategies and best practices:

• Find Common Ground: Finding common ground as key, recognizing that bottlenecks will always arise, and a shared understanding is necessary to move forward. This implies investing time in mutual education and understanding each other's perspectives.
• Active Engagement ("Get Your Hands Dirty"): "Get your hands dirty" suggests a proactive, hands-on approach to understanding the work, rather than observing from a distance.
• Be Flexible and Adaptive: Being flexible is highly recommended. It is important to be able to recognize when things need to change and more importantly, accept change. This allows a collaboration to navigate unforeseen challenges and evolve.
• Maintain Focus on Key Deliverables: Keep eyes on key deliverables and goals. It is sometimes useful to focus on bare-minimum to achieve success. This helps to maintain momentum and ensure progress toward defined objectives.
• Periodic Review and Refocusing: Establish a practice of taking a step back and view where things are and where things are going and what steps are needed for the next success. This ensures the team remains aligned and motivated.
• Regular In-Person Interaction (especially for remote teams): Face-to-face interaction has been found to build and maintain strong collaborative ties, particularly for remote teams, in addition to virtual meetings.
• Prioritize Domain Science Need: Keep in mind that SciDAC is about advancing science. Consequently, what's important is what is important to the domain-science side, even if it may be as simple as making software better. This ensures the work remains relevant and impactful to the primary scientific questions. This can lead to longer and more sustainable collaborations.
• Utilize Common Communication and Writing Tools: Take advantage of communication and writing tools to enhance collaborations. Cloud storage can be used for sharing data and codes. Slack messaging is a great way to ask quick questions. Overleaf is often used for LaTeX/writing. Large Language Models can make it easier to ask simple questions and get a common language. These tools facilitate seamless information exchange and joint work.

Q: What are the pitfalls to avoid when engaging in multidisciplinary collaborations?

• “Throw Codes over the Fence”: Throwing a code over the fence and simply asking others to work on it is never a good way to start a collaboration.
• Bottlenecks and Lack of Common Ground: Not establishing a common ground can lead to stalled progress, particularly when hitting a bottleneck.
• Ignoring Differences Between Communities: Communities (such as universities and national laboratories, or CS/AM researchers and DS investigators) are often different and can have different perspectives. Failure to recognize and navigate these inherent differences can lead to frictions and misunderstandings.
• People Being Spread Too Thin: When team members are overcommitted, their ability to contribute effectively and consistently to the collaboration is severely hampered, resulting in lack of cohesion and progress.
• Lack of Active Engagement ("Not Getting Your Hands Dirty"): Passive participation— where individuals remain detached from the practical aspects of the collaboration— can be counter productive. Look for ways to engage those individuals.
• Goals Changing Constantly : Constantly shifting objectives can destabilize efforts and lead to wasted work. Adjusting goals on an annual basis may be preferable.
• Rigidity and Inability to Accept Change: Being rigid or unwilling to adapt, when necessary, can stifle progress. Try to be flexible.
• Neglecting In-Person Interaction (for remote teams): An over-reliance on purely virtual communication can erode team cohesion. Regular in-person meetings for remote teams are highly recommended.

Last updated: 04 August 2026