Future of Knowledge Synthesis
Updated: Oct 27, 2021
We have so many tools for finding good research papers, organize our notes and improve our writing, and yet, where are the tools for synthesizing our thoughts?
Note-taking apps have existed for decades now in the form of a traditional file cabinet system of drawers and folders. It is only recently that we see an explosion of apps having a more bottom-up architecture (Roam Research, Logseq, Athens, Tiddlywiki etc)
Even though tools of thought have been as old as computing itself, knowledge synthesis is still very much an open problem. (Chan, 2020). And with the advent of collaborative sense-making through such platforms, it opens up wider possibilities for knowledge workers.
We might have already witnessed the potential of knowledge graphs through the trivial Wikipedia game, waltzing across search topics as diverse as bolshevism to celestial mechanics to the Dunbar number. But this could be so much more.
Imagine the potential of a cancer researcher finding an interesting query result of carboplatin to oxaliplatin without reference to melatonin. Such a quick query search at the speed of thought leading to a cure for cancer might soon be a reality with the rise of such collective knowledge graphs.
This has been my primary motivator behind my interest in this research topic and these are my initial explorations along with the guiding research topics. Playing along the horticultural metaphors popularized by Maggie Appleton, I will try to bring more clarity over these questions, in a less performative fashion than a blog, in the form of a garden that is slowly being pruned and nurtured to let it grow from a seedling to a budding, which could soon become something more evergreen and permanent.
What sort of implicit metadata could you capture from people's behaviours?
We suffer from a heavy recency bias in most of our current social platforms. We tend to see a surplus of noise in such 'chaos streams'. However, when we look at knowledge graphs, the intent aligns more towards building curated collected artefacts that are pruned and tended over years.
And as such, the metadata of the chaos streams such as publication date and time become irrelevant as we are not dealing with ephemera here. The objective is more oriented towards sense-making and truth-seeking, the sense of 'connectedness' being the underlying thread (pun intended).
Posts are connected to other posts through related themes, topics and shared context providing important metadata towards the proximity in sense-making. (Appleton, 2020)
Another possible way to imagine such a collective knowledge graph could be based on questions and their relative importance as quoted by Ryan Muller. The sole person involved in the siloed activity of note-taking is slowly transformed into a multi-person institute revolving around formulated questions of interest (popularly called Zettelquestions).
Andy Matuschak talks about how mnemonic mediums are crucial for structuring thought into a knowledge graph, as understanding becomes a key point in this direction. In this case, the metadata of increasing importance might include sources read, relevant notes, spaced repetition scores (SRS) etc. In a way, retaining important information over longer periods becomes a crucial way to detect legible signals in such question-centred networks helping rank posts based on relative importance. (Matuschak, 2020)
For a curious person interested in discovering things about the world and how it works, this is the greatest time to be alive. Podcasts, online education, courses, Youtube, Wikipedia are all such incredible sources of learning. And yet, the parameters of analysis for the existing social networks such as impressions, click-through rates, eyeball time has led to more harm than good.
We have become analogous to the rat which when asked to choose between food and pedal, we choose the pedal. The dopamine hit through the game has become so addictive that we wouldn't even mind the starvation.
The existing social networks are designed towards engagement leading to distraction, outcry and even addiction. But what if the parameter is shifted to more healthy forms such as fulfilment, rather than engagement? We certainly do have internal goal-seeking mechanisms. What would happen if such interfaces cater to such needs?
An example of a 'Fulfilling' Twitter would be when the algorithm that dictates the posts are based on personal values and goals which are set in advance. In a similar sense, could such collective knowledge graphs cater to fulfilment through sense-making?
What user behaviours are people doing already that specifies structure without that being instantiated into a structure?
The cyclical dance between convergence and divergence is an overarching structure often found in creative processes while synthesizing ideas.
In the popular Design Thinking framework, the designer goes through various stages of the process ranging from Designing the Right Thing (Empathize, Define, Ideate) and to Designing Things Right (Prototype, Test). Convergence and Divergence go hand in hand in the form of a double diamond to finally arrive at the solution. (Tschimmel, 2012)
In the divergence phase, methods could range anywhere from active/passive searching, collaborating, and socialising (Herring, Jones, & Bailey, 2009) to very formal procedures like brainstorming and morphological analysis. In the convergence phase, they use a wide variety of methods such as matrixes, argument flow maps and so on to arrive at the outcome. And both of these happen in multiple helical spirals to arrive at the final synthesis.
In this way, the person inadvertently follows a structure without being instantiated into a structure during knowledge synthesis.
There have been various developments on formalized data structures for collaborative knowledge synthesis.
Joel Chan explains that formal structures aren't just for show, but unlock powerful forms of reasoning. As popularized by Marshall McLuhan, the 'Medium is the message'. There have been various examples in the past which include conceptual combination, analogy, casual reasoning etc.
As these logical formalized structures change, the data structure's elemental representation might also differ, creating new types of tools for thought.
How do you create powerful interfaces for structuring thought into a knowledge graph?
Having used Roam for more than a year, and have used Evernote for 5 years, I have gained some insights into the adoption of top-down and bottom-up note-taking tools for knowledge management.
One such observation is the over-emphasis of tools over workflows in the productivity-porn community. The tool certainly doesn't exist in siloes, and the context and workflows become even more important when it comes to structuring thought.
As a novel environment for exploration, it becomes quite useful to start with some guiding metaphors and explore possibilities of creating powerful interfaces around such metaphors. Some of the metaphors which are currently explored here - garden meandering, mind palaces and visual diagrams.
Meandering in a garden
Web 2.0 got formalized with the sweeping away of streams - where information collapsed into a single track timeline of events. The conversational feed design of email inboxes, group chats and social platforms were well designed to trap the immediate thoughts that rush within a few moments, surfacing all the zeitgeisty ideas of the past 24 hours.
Compare this to meandering, where you are slowly making your trail across a garden pathway. The garden explorer follows any trail through the content, rather than being dumped into the most recent feed.
At the moment, these digital gardens are all solo affairs. The future of gardening could soon be multi-player as an enthusiastic community of developers and designers are already trying to set it up.
Digital comes with a fair share of advantages providing almost-infinite memory and computation.
Even though writing with pen on paper might seem so Luddite and anachronistic, a key feature in which this medium shines is retrievability. Although we might have some advancements in SRS to retrieve digital information better, the greater problem comes with having a digital app to combat a digital problem.
There could be a reason why paperback books are better than Kindle for information retrieval.
In the book, Moonwalking with Einstein, Joshua Foer expands his memory retrieval limits and becomes an American Memory Champion and a record holder of speed cards. (Foer, 2012) To become such a champion, Joshua used Memory palaces to combine both visual memory and spatial navigation in a unique way for faster retrieval. (Fassbender, 2006)
Tying concepts to visually stimulating objects (blue banana, purple cow etc) and then spatially fixing them in a place helps them remember not just their grocery list, but even 1000, maybe 10,000 digits after the decimal place of irrational digits such as pi.
Could it be the spatial nature of our memory, that makes us remember physical books more than the ones in Kindle? In the mind palace analogy, the information is given a spatial representation in a mental environment to make retrieval better.
What if this could be aptly used to build collective knowledge graphs where you can meander but not get lost?
Imagine navigating through a 'town map' for 'places'. Similar to how in a digital garden you are taking a stroll through the articles written by the author, and expanding your horizons. A version of such pseudo-physicality is already being explored by the digital gardeners in the form of central indexes with examples of sections that include essays (distillery), artworks (museum), newsletters and podcasts (planetarium), brain (library) etc
Affordances for augmenting existing streams of consciousness
In a collaborative online reading process undertaken by Finnish students, the usage of a web-based representation tool allowed for the effective synthesis of information from the internet to compose well-formulated essays (Killi and Lieu, 2019). These patterns suggest that additional work with representational tools such as argument graphs during online reading could yield more promising results of synthesis.
Nearly half of our human brain is directly or indirectly related to the processing of visual information. However, most of our knowledge synthesis work has been primarily through the usage of textual stimuli.
There has been an increasing list of evidence that have shown improvements in the knowledge synthesis process using visual stimuli which can not be discounted. (Goncalves et al, 2016)
It's an interesting direction to notice such knowledge graphs encompassing multi-media in a more retrievable fashion to facilitate more creative connections.
Chan, J. (2020, April 30). Roam's superpower for knowledge synthesis: Incremental formalization. RoamBrain.com. https://roambrain.com/knowledge-synthesis/.
Appleton, M. (n.d.). A Brief History & Ethos of the Digital Garden. Maggie Appleton. https://maggieappleton.com/garden-history.
Matuschak, A. (1970, January 1). How to write good prompts. How to write good prompts: using spaced repetition to create understanding. https://andymatuschak.org/prompts/.
Tschimmel, K. (2012). Design Thinking as an effective Toolkit for Innovation. In ISPIM Conference Proceedings (p. 1). The International Society for Professional Innovation Management (ISPIM).
Foer, J. (2012). Moonwalking with Einstein: The art and science of remembering everything. Penguin.
Fassbender, E., & Heiden, W. (2006). The virtual memory palace. Journal of Computational Information Systems, 2(1), 457-464.
Kiili, C., & Leu, D. J. (2019). Exploring the collaborative synthesis of information during online reading. Computers in Human Behavior, 95, 146–157. https://doi.org/10.1016/j.chb.2019.01.033
Gonçalves, M., Cardoso, C., & Badke-Schaub, P. (2016). Inspiration choices that matter: The selection of external stimuli during ideation. Design Science, 2, e10. https://doi.org/10.1017/dsj.2016.10