🪴 Scaling Synthesis

Search IconIcon to open search

Q- What is synthesis

Last updated April 29, 2022

Authored By:: P- Joel Chan

We use the term “synthesis” in this project to refer to a particular form of sensemaking, where the inputs are practical, scientific, or scholarly theories, findings, and data, and the output is a coherent intellectual whole that can guide decision-making for complex problems with no clear playbook to follow, and/or spur conceptual innovation to advance the state of knowledge on a problem.

Synthesis may be supported by and manifested in a variety of forms, such as a theory, an effective systematic or integrative literature review, a causal model, a cogent research proposal or problem formulation, or model of a design space, among others. The core commmon intuition is the creation of a novel conceptual whole that is greater than the sum of its parts R- Types of synthesis and their criteria.

To illustrate the difference between a “mere assemblage” and a true synthesis, consider the difference between these two hypothetical ‘summaries" of a set of findings:

ExampleMere assemblageSynthesis
Related work for systems paperThere are a few ways to solve problem X. One is [1,2]. People have also tried [3,4,5]. Finally, others have tried [6,7,8]. In this paper, we introduce our solution Y that solves problem X.The earliest attempts to solve problem X took the angle of [1,2]. But it has limitations ABC. For example, [1] showed that… To address limitations ABC, others have explored the possibility of [3,4,5], which makes key modifications XYZ to . This yields some gains under settings DD, but doesn’t work in settings EE. The overarching open problem seems to be the tradeoff between XXX (common in settings EE) and the side effect BB of . To address this open problem, researchers are exploring [6,7,8]. has shown some promise, yielding X% gains in settings EE; but a remaining open problem is YY. In this paper, we build on these efforts by addressing this key limitation.
Darwin’s theory of evolutionSpecies vary: some variations are bad, and some help with survival. Species struggle to survive. Some, but not all, organisms pass on new offspring.Species struggle to survive. Species also vary, and some variations are good and some are bad for survival. Therefore, one precondition for species to survive and pass on offspring is by having or inheriting beneficial variations. This variation and selection process explains how we get the diversity of species we see today.

In both examples, mere assemblage simply lists an undifferentiated series of findings in succession. In contrast, a synthesis of the hypothetical set of findings in both cases, weaves them together into something greater than the sum of the parts: in the system paper example, a theory of change for what the most fruitful angle of attack is for making progress on a problem; in the latter example, a full-blown causal theory of the origin of species!

You would be surprised at how common this “phone book” mere assemblage pattern is in academic literature reviews, from doctoral dissertations, even ones that pass! (R- Making the Implicit Explicit, R- Investigating PhD thesis examination reports, R- Scholars Before Researchers), and even many published papers (R- A Troubleshooters Checklist for Prospective Authors, R- Generating research questions through problematization, R- Theory Before the Test, R- A decade of theory as reflected in Psychological Science, R- Using systematic reviews to inform NIHR HTA trial planning and design, R- Cochrane and non-Cochrane systematic reviews). This is why we use the term synthesis instead of the more familiar “literature review”: in today’s conventions, only effective literature reviews are actually synthesis! We think this is harmful for scientific progress and innovation, and are therefore motivated to build new knowledge infrastructures that could make effective synthesis less painful and more commonplace, such as a decentralized discourse graph.

Synthesis is necessary for innovation, and as outlined in C- Discourse graphs could significantly accelerate human synthesis work, a decentralized discourse graph can dramatically reduce the work required to achieve synthesis.