Tinderbox Forum

On knowledge management, Tinderbox in relation to other apps

Amen. I don’t think either developers of ‘KM’ apps nor their users are doing anything wrong. But, in the background is a naïvety about what ‘the computer’ can and can’t do for us. At the same times our personalties and types of work differ more than we like to acknowledge; one person’s beautiful system is another’s complex nightmare.

The fallacy of some of these knowledge tools is that, despite their ease of capture of information, the act of its capture alone is simply no guarantee of insight. Keyword matching works best for terms with precise meaning (i.e. specific context of use) and no alternate meaning. Automatic linking merely implies presence (mistaken presence). The strictures of such linking (and problems over duplicate note names) force an unrealised and likely unintended controlled vocabulary on the dataset even where not desirable. Back links are fine as a navigational conduit, but if the intention of the linking (less possible if fully automated—without any logic in that function) is not captured it is more a simple graph than a skein of knowledge.

My takeaway from this is use the tool that works for you. But, be careful mistaking ease (lack of ‘friction’) for depth and insight. Nor does that reply the reverse. Harder is not better. Rather, it is intention and a willingness to allow emergent structure—instead of forcing/imposing it—that tends to offer the most reward.

Of course, if you are mapping a syllabus of fixed process then you do want structure. Tinderbox supports in both these roles, and other tools are available.

When comparing app A (that we know well and have used for some time) with app B (where we’ve little experience or time in use) it is necessary to be careful to avoid skewed comparisons. Invariably we compare B to A and judge B by its comparison to A’s features. That’s great if B is meant simply to replace A, but mainly indicates how easy it might be to learn to use B. I’m as prone to this flawed approach as any. I try instead to ask what B does that A doesn’t, or does in a less easy/productive way.

Also massively overlooked is import and export. Unless you know for certain that info will never leave your KM app except by reading in the app, import and export are more important than you might assume. A disproportionate amount of time in the forums over the years has been spent trying to bridge the inadequacies of import or export in other apps (not that Tinderbox doesn’t throw up the odd head-scratcher). Otherwise a KM tool is just a ‘roach motel’ for you information (i.e. information “checks in but never leaves”).

I’m an ‘untidy desk, untidy mind’ person - not a very linear thinker. Though it has taken a long time to learn the deeper parts of Tinderbox (I still am!) the appeal is that it is not pre-structured, allowing room to explore an idea or goal. I’m not surprised—and happy to read above—that some others work here also without a fixed system. That does not imply a lack of intent or purpose. Rather, it acknowledges that the path to the goal is not always self evident at start; there may be blind allies on the way, with no straight path to the goal. Few tools support open thinking and exploration in this way way. Would that there were more! However, it is worth noting such exploratory tools do reward relaxing the overly linear per-structuring we are taught at school.


So many quotes in your post that I love, but this one made me laugh. :slight_smile:

1 Like

I’m that kind of thinker as well, but also I’ve learned to be productive by imposing constraints and pattern on how I work. One of. the benefits of Tinderbox’s unstructured base is that I can put any kind of structure on top of it. Over the years, I’ve learned I need several levels of imposed organization as project and folder to help focus my habit of getting too far afield. So I tend to build focused files in Tinderbox and impose some organization once I can’t take the whole thing in at a glance.

For me, these new link based note capture systems rapidly turn into the unusable rats nest of links that some of my big Tinderbox files eventually turned into. When it gets to be a junk drawer, it’s time to do a thorough cleaning and organization. Or a dump in the trash to start over, which often is the easiest way forward.

1 Like

I think frictionless is an important concept that is applicable as a general best practice. It is something that can be applied to all facets of life. It is an aesthetic to strive for. In software, it keeps things approachable, useable, and enjoyable. I would separate this from the notion of unstructured ingestion in a knowledge management system. I see no reason why you could not have a system that, at ingestion, forces a structure predefined by the user that is frictionless.

When I analyze why I use Obsidian heavily in my day-to-day note taking, I come to the realization that what I need at those moments in time is something that I don’t have to think about, a place where I can trust that I will be able to pull things together later for further processing and thought. My use case is not academic, it is work related. I need to dump notes somewhere quickly and get back to herding cats. I think I could do this in Tinderbox, but it wouldn’t be as effortless at the time of ingestion. Now, when it comes to deep analysis, I think importing the relevant notes into Tinderbox is very powerful. Tinderbox not only allows you to develop and apply sophisticated structure and automatic manipulations, it also allows you to see your notes in myriad visual ways. It is simply unparalleled as a “note thinking tool”.

So, my thoughts as to how the pieces fit together for me currently are:

Devonthink - source/artifact storage and retrieval
Obsidian - capture of notes in realtime, quick linking
Tinderbox - deep thought and processing

I find this topic fascinating.

Great discussion about knowledge management here. I’m actually still using Roam, Tinderbox, Evernote, The Brain and MaxQDA in parallel - and find each one has subtle advantages for different kinds of task which have meant I have not been able to alight on a consolidated, single KM platform. I’m waiting for some special features that I think will make all the difference: 1. an automated ability for the app to recommend to me potential knowledge management actions that would be conveyed to me conversationally, upon demand (perhaps via an interface such as replika.ai) - this might, for example, reveal connections between items that I’d overlooked, suggest actions I should take based on inferences about the significance to me of recognised connections, suggest insights or KM strategies I should consider by comparing what data I’ve collated and how I work with my information to the methods used by other users of the software that I am willing to accept into my knowledge circle, &etc; 2. an ability to navigate through information in a compelling, immersive, fly-through manner (additionally capable of taking advantage of an AR/VR headset if present); 3. an ability to build a model of my personal outlook (“Weltanschau”) through machine empathy so that as I collate more information, the application is able to populate an anterior sense of my wants, needs, interests, behaviors, habits etc so that it can potentially, with my approval, build predictions about things it could do to add value to my life as an agent/negotiator or just a kind of background consulere (this sounds like Alexa but could be so much better if capable of receiving the right training, with the right memory to support spontaneous suggestions along a corridor of persistent context…). Anyway, this is all fairly trite but I wonder how many years off this is and whether the development roadmaps of the programs I use are actively pursuing such goals. It would be marvellous if some of this could turn up in the next version of TB :smile:

1 Like

I am pretty sure it will be in v.9. Kidding aside, I think these are the places AI will go -some kind of human-machine collaboration.

I’m always mystified that we seem to want the computer (‘AI’) to do things humans are still better at—associative thinking, understanding nuance in written prose—whilst not leveraging the capabilities it is good at—fixed patterns, scheduling, detail replication. We keep having cycles where we believe if we only just tip enough data (notes, records, whatever) into the system that a magical enlightenment will appear.

edit: some text got posted in the wrong thread. Sorry for confusion.

Well, there’s nothing wrong with pushing the envelope of what computers can help with!

What’s particularly interesting right now is to imagine how the machine, brimming with suggestions — many of them bad! — can be helpful. We’re beginning to do extraordinary things: for example, the automatic recognition of places and organizations that Tinderbox does today is really remarkable. But it’s easy to be overwhelmed with “helpful suggestions” you don’t really want!

I did a paper on this at ACM Hypertext 2020: BAD CHARACTER: who do we want our hypertexts to be?

Yes, I remember Clippy too :scream:

1 Like


Working on a links training this week. How does linking play into your knowledge management efforts? What does a link association visually or structurally give you that non-linking does not? Does link direction matter? Link format (width, style, endpoints, color)…?

What else?

As for a how-to? What would you like to see in a Tinderbox training video on linking?

Michael, this sounds like a wonderful complement for the issues I am grappling with that I briefly mentioned in the last meetup (shepherding manuscripts through a review process).

What I hope to achieve with linking is to, well, link two or three things that are conceptually different yet have a relationship. And their distinctness is the reason I do not want to roll all aspects (all Displayed Attributes, or (in database lingo) all fields / variables) into one e.g. prototype.

Because (in my case) I have many reviewers per area of specialisation or a reviewer (over time) dealing with many manuscripts.

But how do I best achieve that? Do I use a primary key for the link (which will make things perhaps less readable and require a special template to generate a “report”) or do I use plain text (say like names) to do the linking (but what if I have several David Millers that I need to distinguish)?

Looking forward to your video, as always! Many thanks for your work.

Don’t overlook link types - you can define you own. The link type is what appears as a link label in Map view, and so often mistaken for a simple label. But it has a richer semantic meaning. Action code queries (e.g. linkedTo) can be constrained to only match certain link types. Plus it is possible to link two objects multiple times, each with different link types (which functionally equates to two items links by multiple link types).

Don’t think it too fixed terms about that. A note—or prototype’s—Displayed Attributes are simply a list in the note’s $DisplayedAttributes attribute. So you can easily swap different lists of Displayed Attributes into $DisplayedAttributes depending what you are doing. prototypes help with that, e.g. you can instantly change the Displayed Attributes for 100s of notes whilst working on an aspect of work that is helped by seeing different attributes as Displayed Attributes.

Well every per-reviewer note could link to all per-paper notes where they are a reviewer, with a link type ‘reviewed’. Or if you need more specificity ‘reviewer 1’, ‘reviewer 2’, etc. Or, ‘specialist reviewer’ if they have special role in the process.

Note that the latter is putting metadata into a link (as the link type value) that is then only stored in that link. You may find it useful to store associated date (e.g. for a reviewer, the papers reviewed) as a list of $ID numbers. Those never change and are always unique.

A point sometimes lost for those who start and stay in map view. Links are more than just lines with optional text labels. Used deliberately they also hold semantic meaning that can be queried leveraged.

They are yet another way to expand beyond constantly using an agent and .contains() on $Name or $Text. To mis-quote a song:

If you valued it so much you should have put some metadata on it

Recall spending ages finding that relationship of interest in one of the aliases inside a now deleted agent only to have to repeat the process again? Metadata, i.e. attribute values or link types, capture that relationship and create addressability for otherwise arbitrary seeming interrelationships of notes. This is something that isn’t always self-evident to all new users so it would be good to draw out in this or a later lesson.

1 Like

Linking provides 2 benefits over alternatives (such as using tags):

  • Easy navigation from “linker” to “linkee” (and back)
  • Option to represent the relationship visually

Methods like tags have their own benefits (easy searching, filtering and sorting) and so I see links as adding a set of alternative ways of looking at relationships.

It can, if you need it to, but it doesn’t have to. If I’m using links to indicate progression, then direction matters. If A causes or leads to or implies B, then I’d want the link to show that. If A and B are members of a set of related items, then I may well want to avoid direction indicators.

I have no strong opinion about these

1 Like

It’s probably worth noting that as the Hyperbolic view matures (i.e. (if) it gets link type /scope filtering) then links will gain even more usefulness.

Even if links are used, storing some metadata in notes as well can be useful, especially if using action code to make/remove links or using link actions. It gives some redundancy in case of user error.



One vote for t-shirt.


I completely agree with Mark’s sentiments above. Another way of making the point raised is this.

“Every note has every attribute, whether or not the attribute is visible.”

TBX TIP: A Tinderbox note is like a stem cell. Cells are “the structural, functional, and biological units of all living beings.” Notes in Tinderbox, like cells in the body, are the basis of structure, function, and bring knowledge to life. A stem cell is “an undifferentiated cell of a multicellular organism which is capable of giving rise to indefinitely more cells of the same type, and from which certain other kinds of cell arise by differentiation;” this is exactly what a Tinderbox note is. What differentiates TBX notes from each other is the value in attributes. So, in Tinderbox, like a stem cell in the body, a note has unlimited potential. What you do with it determines its future.

Remember: As Mark points out, $DisplayedAttributes, what you see in the “Text” panel is nothing more than $Name, the value of $Displayed Attributes (if not collapsed), and $Text. So what you choose to show in $DisplayedAttributes does not make it significantly different than any other note, this value is just one of a sea of attributes.

1 Like