From the risk of stagnation to a vision to metascience. A bird’s eye view of some interesting ideas and initiatives about science and the future of science in 2022.

Risk Aversion Is Ruining Science

To nurture scientific discovery, universities and grant institutions must do a better job of rewarding failure,

Paul M. Sutter, 27 April 2022, Undark

The culture that celebrates, supports, and rewards the audacious mental daring that is the hallmark of science is at risk of collapsing under a mountain of cautious, risk-averse, incurious advancement that seeks merely to win grants and peer approval.

Some risk-aversion is to be expected in science. As research fields mature and scientists pick off more of the low-hanging fruit, the problems become harder, requiring more people and more resources to solve. It’s also easy to fall into a trap of conformity. Graduate students work on the problems that their advisers find interesting.

The realities of the current grant funding climate play a role, too. It’s becoming increasingly difficult to get a federal grant.

This culture of risk-aversion is putting science itself at risk. To prevent science from becoming yet another bureaucracy that exists only to perpetuate itself, scientists have to begin by making changes at a cultural level.

The first step is to reward risk. We can incentivize risk

• by hiring and promoting junior scientists who do something new, even if they weren’t lucky enough to have it pay off.
• in the grant proposal process, with “support for high-risk proposals with potential for transformative advances”
• But perhaps just as important, scientists need to manage the expectations of the public and policymakers.

Here is another idea:

Fund-by-variance: Instead of funding grants that get the highest average score from reviewers, a funder should use the variance (or kurtosis or some similar measurement of disagreement) in reviewer scores as a primary signal: only fund things that are highly polarizing (some people love it, some people hate it). One thesis to support such a program is that you may prefer to fund projects with a modest chance of outlier success over projects with a high chance of modest success. An alternate thesis is that you should aspire to fund things only you would fund, and so should look for signal to that end: projects everyone agrees are good will certainly get funded elsewhere. And if you merely fund what everyone else is funding, then you have little marginal impact

(Michael Nielsen and Kanjun Qi, Fund-by-variance, quoted below)

Science Has a Communication Problem

And a Connection Problem, too. The public needs to see scientists as people, rather than simply sources of information. What barriers stand in the way?

Paul M. Sutter, 20 October 2022, Undark

U.S. adults that have a “fair amount” or a “great deal” of trust in scientists fell from 87 to 77 percent in just the past two years.

Research has shown that simply giving the public more information isn’t the best way to correct this. It’s crucial that scientists find ways to communicate more effectively and directly with the public, so the public can have access to the minds, and hearts, of scientists. In other words, they need to see scientists as people whom they can empathize with and learn to trust. To enable this, scientific institutions must support those endeavors via media training and institutional incentives which are currently lacking in the academic landscape.

While the majority of scientists use social media, many of them use those platforms to connect with each other, rather than the public. Scientists should use that newfound voice to speak directly to the public, removing any barriers and distortions placed by gatekeepers.

Can a new approach to funding scientific research unlock innovation?

How we fund research is stifling creativity. Here’s one potential fix.The Arc Institute

Kelsey Piper, 18 December 2021, Vox

Ask a bunch of scientists what’s wrong with their field and one thing nearly all of them will name is the funding process.

By some estimates, many top researchers spend 50 percent of their time writing grants. Interdisciplinary research is less likely to get funding, meaning critical kinds of research don’t get done. And scientists argue that the constant fighting for funding undermines good work by encouraging researchers to overpromise and engage in questionable practices, overincentivizing publication in top journals, disincentivizing replications of existing work, and stifling creativity and intellectual risk-taking.

A wide range of ideas have been aired for how to fix the scientific grant process, from lotteries to limiting applications to one page. There have been private attempts to do better — like FastGrants, which aimed to get out Covid-19 research money in 48 hours instead of weeks or months, and which has moved more than $50 million to date.

Arc Institute is an institutional experiment in how science is conducted and funded. Researchers get eight-year grants to do whatever they want, instead of three-year grants tied to a specific project. The basic thesis behind Arc is that there are a couple of ingredients that haven’t been combined before:

• Full, “hard money” support for investigators, so that they can pursue curiosity-driven research programs in unfettered fashion.
• Close partnership with a number of major research universities, including top graduate programs.
• First-class investment in technology development to support research programs, especially genome engineering and software engineering.
• Long-term career paths both for researchers focused on technology development and for Core Investigators.
• Physical colocation of researchers.

Arc was started by Silvana Konermann, Patrick Hsu, and Patrick Collison. Founding donors include Vitalik Buterin, Patrick Collison, among others.

Large language models will change science

Patrick Mineault, 30 May 2022, Xcorr

What if every single one of us had infinite access to world experts in any field of field of study? They could recommend papers to read, based on:

• relevance
• accuracy
• importance & novelty

We’re not far off from that world. Advances in natural language processing (NLP) and large language models (LLMs) mean that it’s becoming possible for machines to understand natural language, including the prose in scientific papers. We have only started to see these advances deployed, but my bet is that it will change how we consume and create science over the next few years. Let me recap the progress in NLP over the last 5 years and paint a vision of scientific natural language processing in the near future.

The road ahead.

How to Build a GPT-3 for Science

Why don’t we have a DALL-E or GPT-3 for science?

Josh Nicholson, 18 August 2022, Future

Can you imagine if a researcher could propose an experiment and an AI model could instantly tell them if it had been done before (and better yet, give them the result)? Why don’t we have a DALL-E or GPT-3 for science? The reason is that although scientific research is the world’s most valuable content, it is also the world’s least accessible and understandable content.

Because scientific papers are not easily accessible, we can’t easily use the data to train generative models like GPT-3 or DALL-E. We need to treat scientific publications as substrates to be combined and analyzed at scale. Once we remove the barriers, we will be able to use science to feed data-hungry generative AI models.

• Decompose papers into their minimal components
• Access all the components
• Connect the components and define relationships
• Use relational data to train AI models

Fortunately, papers are becoming more open and machines are becoming more powerful. Liberating the world’s scientific knowledge from the twin barriers of accessibility and understandability will help drive the transition from a web focused on clicks, views, likes, and attention to one focused on evidence, data, and veracity. Pharma is clearly incentivized to bring this to fruition, hence the growing number of startups identifying potential drug targets using AI — but I believe the public, governments, and anyone using Google might be willing to forgo free searches in an effort for trust and time-saving. The world desperately needs such a system, and it needs it fast.

SciSci: Science of science

Science can be seen as an expanding and evolving network of ideas, scholars, and papers. The science of science⁽¹⁾ (SciSci) searches for universal and domain-specific laws underlying the structure and dynamics of science.

Fortunato, Santo, Carl T. Bergstrom, Katy Börner, James A. Evans, Dirk Helbing, Staša Milojević, Alexander M. Petersen, et al. ‘Science of Science’. Science 359, no. 6379 (2 March 2018): eaao0185. https://doi.org/10.1126/science.aao0185. 2 Mar 2018, Science

The increasing availability of digital data on scholarly inputs and outputs—from research funding, productivity, and collaboration to paper citations and scientist mobility—offers unprecedented opportunities to explore the structure and evolution of science. The science of science (SciSci) offers a quantitative understanding of the interactions among scientific agents across diverse geographic and temporal scales: It provides insights into the conditions underlying creativity and the genesis of scientific discovery, with the ultimate goal of developing tools and policies that have the potential to accelerate science. In the past decade, SciSci has benefited from an influx of natural, computational, and social scientists who together have developed big data–based capabilities for empirical analysis and generative modeling that capture the unfolding of science, its institutions, and its workforce. The value proposition of SciSci is that with a deeper understanding of the factors that drive successful science, we can more effectively address environmental, societal, and technological problems.

A Vision of Metascience

An Engine of Improvement for the Social Processes of Science

Michael Nielsen and Kanjun Qiu, October 18, 2022, The Science++ Project

In this essay we sketch a vision of how the social processes of science may be rapidly improved. In this vision, metascience plays a key role: it deepens our understanding of which social processes best support discovery; that understanding can then help drive change. We introduce the notion of a metascience entrepreneur, a person seeking to achieve a scalable improvement in the social processes of science.

Authors argue that:

1. metascience is an imaginative design practice, exploring an enormous design space for social processes;
2. that exploration aims to find new social processes which unlock latent potential for discovery;
3. decentralized change must be possible, so outsiders with superior ideas can’t be blocked by established power centers;
4. ideally, change would align with what is best for science and for humanity, not merely what is fashionable, politically popular, or media-friendly;
5. the net result would be a far more structurally diverse set of environments for doing science; and
6. this would enable crucial types of work difficult or impossible within existing environments.

For this vision to succeed metascience must develop and intertwine three elements:

• an imaginative design practice
• an entrepreneurial discipline,
• and a research field.

Overall, it is a vision in which metascience is an engine of improvement for the social processes and ultimately the culture of science.

A few specific examples of unusual social processes that could be (or are being) trialled today by adventurous funders or research organizations:

• Fund-by-variance
• Century Grant Program
• Tenure insurance
• Failure audit
• Acquisition pipeline for research institutes
• Pull immigration programs
• Open Source Institute
• Institute for Traveling Scientists
• Public hall of shame / anti-portfolio
• Interdisciplinary Institute
• At-the-Bench Fellowship
• Printing press for funders
• Excitement quotient for funders

_____________________

(1) Metascience is the use of scientific methodology to study science itself. It is also known as meta research, research on research and the science of science.

Featured Image: Mike MacKenzie, Artificial Intelligence & AI & Machine Learning