Information Architecture fundamentals

Jorge Arango is offering a course about Information Architecture, and for me, considering it was rather appetizing. So I asked myself how much I already know about that topic. Off the top of my head, here is what I found fundamental in my practice.

1. Push vs. pull
2. Link ‘gender’ (down vs. across)
3. Customer vs. provider view
4. Addresses vs. labels
5. Click-haters vs. scroll-haters
6. Search vs. browse
7. Categories vs. tags
8. Teaser vs. summary
9. Bold vs. italic

1. Push vs. Pull

As information providers, we must be honest with ourselves and be clear about what is the predominant reason for posting some information: do we want a mature customer or reader to pick what they desire to know, or do we want to ‘sell’ them some bargain, or evangelism, or shop-warmer, or a breaking warning?

2. Link ‘gender’

For each link on every page, we must always be aware whether the link is a hierarchical one within a classification tree, or a cross-reference/ shortcut/ “see also” link. The difference is as fundamental as distinguishing a bearing wall from a divider wall — in the building architecture comparison that @jarango uses to explain IA.

3. Customer vs. provider view

As the first webmaster of my university, I tried to explain to the various units who wanted a web presence that they are the ‘decorators’ of the ‘shop window’, with a different access from behind, and with different interests and concepts. But today, many sites just replicate their internal organization and ignore that customers have different terms in mind.

4. Addresses vs. labels

From the provider’s view, addresses (URLs) looking like file and folder names are great and make administering easy, but the labels to be clicked deserve some more thought.

I think Postel’s law can be fruitfully applied also to labeling taxonomy: “be conservative in what you send, be liberal in what you accept”. I.e.: use consistent exact labels, but offer plenty of additional variants in alphabetic keyword listings for redirecting. (BTW, please repeat the letter springboard often enough, such that I don’t have to resort to CTRL + F when I want to avoid typing into your search box!)

5. Click-haters vs. scroll-haters

As you might have guessed from my last sentence, I’m a scroll-hater. But the point is: there will always be different types of preferences, and we should never ever try to infer some optimum from our own taste, nor from painstaking user interviews. Click-haters are often the loudest, and accordingly, many UIs overrate hovering and typing.

6. Search vs. browse

This distinction is not only about individual preferences and styles, but also depending on the occasion for one’s visit, and even on a much more fundamental difference: seeking a narrow response or a wider context, looking up a known item or recognizing it only when seeing it.

7. Categories vs. tags

Sometimes, ‘category’ just means that it can be nested while tags can not (e.g. in WordPress), but other times, categories are used as ‘Mutually Exclusive, Collectively Exhaustive (MECE)’, while tags mean that items may have multiple, optional tags. For browsing (6.) in a wide, hierarchical (2.) context, the MECE categories seem to be more appropriate, but for liberal acceptance (4.) of user questions, tags may be more promising.

Unlike the old, browseable card catalogs, digital classification systems offer much more flexible browsing, for example, permutation of index-words, such as swapping of place and time (and person and topic and form) descriptor keywords. Such ‘faceted classification’ is a fascinating alternative to typing into a search box.

8. Teaser vs. summary

Between headline and full-text, there is often a medium-sized ‘description’ or ‘lead-text’ possible. If the user is supposed to stay in their ‘learned helplessness’ (minority) and to earn us clicks, we won’t tell them too much about the full text and offer just a teaser. Otherwise, when we value their time and independence, we might provide a summary or even craft a manual ‘excerpt’, e.g. in WordPress. But mostly, the shorter description is derived just by truncating the long form, which mostly amounts to a teasing click-bait, since the beginning is seldom pregnant. So, the choice is more or less the above question of push or pull (1.).

A related problem is often just negligence: many pages don’t have an HTML ‘title’, which Mac users don’t seem to notice, but users who appreciate a caption on their window or on the taskbar or on a bookmark in their folders, will sadly miss this information type.

9. Bold vs. italic

It is funny that I have never read a plausible advice when to use which of them. They are just seen as different types, or maybe degrees, of highlighting. IMHO, the choice is a bit related to push and pull: Bold is an eye-catcher that helps to quickly find a spot on a page, for intentional, self-directed picking (‘pull’). By contrast, italic is narrower and forces the reader to slowdown, and suggests to stress a word, or directs our attention to a different mode of any kind (quotation, or meta, whatever), i.e., more ‘push’.

Does the text markup belong to literacy or to the arts — probably both camps point to one other. Similarly, I don’t really know what’s the difference between Information Architecture (that can be compared to building architecture which, over here, counts as art), and the other discipline that corresponds to construction engineering (if there is such a discipline, which might have a few simple rules that would rule out the many stupid errors that we encounter each day in our user experience). But I am just a layperson, and maybe my above rules cannot be counted as art, i.e. not as architecture, but just as engineering?

Wanted outcome

Stephen Downes asks us a difficult question:

“what I ask readers to consider is what we want the outcome of an education to be by considering the four points of a pyramid:

a young Marc Andreesen who as a student at a public university was able to develop the first commercial web browser;

the second, the author of the piece, an apparent ‘intellectual’, familiar with the works of Agnes Martin, Claes Oldenburg, Julius Evola and Corey Robin;

the aforementioned billionaire who believes people like him “should get to make decisions to reorder life as we know it without interference from anyone else”;

or the people in small towns who value “the ordinary comforts of kinship, friendship, craft, memory, legend, lore, skills passed down across generations.””

None of the options is ideal, but I won’t duck away and will explain my choice.

1. The student first and foremost represents potential.

2. The ‘intellectual’ in the list is a writer who certainly impresses and outperforms many with his wit and erudition, but I don’t know enough to estimate his concrete impact for us.

3. I have also thought about the desired outcome before, and I want to quote from here:

“According to Stephen Downes, education should help the student to “become the kind of person they want”¹). […] there is the chance that the student themselves readjusts his or her wish about what they want to become. […] if the student’s wish was just to be a rich person without much effort, and now they see that this also entails cultivating recklessness and unsocial skills, which is not what they really want, either?”

OK, if they still do want to become recklessly rich, there is still our chance that they will fail, otherwise we must apply democracy (to neutralize their economic power, the power of the few, built on scarcity, by the political power of the many. I think this aspect of democracy is getting increasingly forgotten, especially if some people think their ‘intellectual’ superpower might be better for the problems than the masses.) So, no to the billionaire with his manifesto.

4. The option of the small town life has a lot of appeal for me, not only because I grew up in a rural place. I remember, the essay title of my final A-levels exam in 1972 was something like “Life off the shelf?” and asked us to compare large city vs. village life; I argued for a smaller town in between.

Later I have come to regard networks where much knowledge (and ethics) is learned “from individuals in one’s close proximity. Via ‘ripple’ effects or, as I expressed it in my first vague post, via contagion. Later I learned that this is compatible with connectivism, see ebb and flow. And it has a lot to do with decentralisation, as opposed to central authorities and templates.” (Copied from this ethics21 post). And I have a lot of respect for indigenous ways of knowing which is similarly learned.

But I am not sure if I would like the hierarchical element suggested by the formulation “passed down across generations”, and the associated stagnation that seems to exclude progress. Also, the context here was about people who don’t leave these places, and I would prefer taking to the road like the journeyman and then come back or not.

So I would aim at option number 1, the promising student, because the potential is important, even if the actual outcome may sometimes be awkward.

Self-incurred minority

For Kant’s 300th birthday today, I looked up some important ideas and their translations. There is one term that seems very topical, but I am not sure if its translation has sufficiently strong connotations: Unmündigkeit — minority, as in this quote:

“Enlightenment is the human being’s emergence from his self-incurred minority.”

(Source: Kant, 1784, An answer to the question: What is enlightenment? in: Practical Philosophy. Ed. and trans. Mary J. Gregor. Cambridge: Cambridge University Press, 1996.)

I wonder how much the term and its opposite remind English-speaking readers just of legal minority/ majority rather than of a much wider context? In German, ‘mündig’ means a lot more than ‘of full age’, and in particular, the absence of a ‘Vormund’ ~ guardian who does ‘bevormunden’ ~ patronize me. (We also feel a connotation with ‘voice’ because of the word component of ‘mund’ = mouth, although this is etymologically unrelated).

As my feed subscribers will know, the topic of patronizing software and user interfaces has engaged me for a long time. Working in IT, I do know what is possible, and I can recognize what are unintended excusable glitches and what is deliberate patronizing. The latter make me increasingly furious.

Now that ‘Tools for thought’ and AI are getting popular, I realize how much the spreading ‘minority’ is voluntarily opted for, and indeed self-incurred.

So, “Sapere aude!” is as very timely advice by Kant, “Have the courage to make use of your own understanding!” (ibid.)

Siemens’ AI optimism

In an impressive TEDx talk, George Siemens explained how he sees AI as an opportunity.

It is not a comfortable opportunity, because it builds on the idea that humans might not be as significant as we thought we were (13:27) which Downes compares with the Copernican revolution. And it comes with force:

“And so AI forces us to look at ourselves in the mirror”

Image CC-BY by Flickr user Tsahi Levent-Levi

But Siemens explains why it is necessary: “On a fundamental level, we have lost a type of core connectedness” (11:02), and his focus on connections is no surprise after his contribution to Connectivism. Here some quotes detailing his observation:

“When I was a child … I had a direct relationship to everything around me.” (09:25)

“I was in a community where everything I did was entangled with nature.” (03:04)

“We live … in a simulated reality today. We have simulated relationships, they’re mediated through a screen. … This technology interface comes with a cost: we are also less healthy” (08:41)

“we continue to add layers that distract us from the world around and so what we emphasize in this space is an isolated layer, through Zoom …” (10:27)

The idea of isolating layers and interfaces resonated well with me, as I have often thought about the interface between us and a new ‘tool’ that often becomes a prosthesis when capitalism seizes it. Then we lose our agency to wield the tool at will in our own way, as if we were disconnected from it by its centralized interface. And the naive willingness to let the babble chatbot think for us, is a particularly ugly example of this enshittification.

So this peak of patronization might indeed be the point where we are startled by the extent of our dependence, and an opportunity to think about the dangerous mixture of capitalism and AI.

Beyond the page

I participated in Stephen Anderson’s Creative Challenge “Make A Zine!”. It was a very useful exercise because (as he says) it “Forces clarity: I know the constraints of only a few panels will force me to distill my thinking on a complex topic down to the essentials”.

Here is my contribution as 8 pages PDF.

Another ChatGPT disappointment

Now we know how OpenAI imagines ‘personal’ AI, and a last hope for a genuine helper tool is gone.

From the examples shown in their announcement (family’s preferences, simplicity preferences, travel plans), we can see that they take two more steps in the wrong direction: again just conversational data rather than unverbalized thought patterns, and central processing rather than on my own device.

I have long questioned if a commercial system would recognize thinking patterns from my own reading and writing behavior, such that I could use it as a cognitive tool. But obviously it just remembers what I narrated to the robot that behaves like an independent actor, and what I told it via the service interface. So it knows what, for example, a neighbor could know about me, but not help me think.

And the suspicion suggests itself that this neighbor knowledge is what is commercially lucrative — despite the affirmation of my being “in control”. Even if the system does nor immediately exploit that knowledge for commercial offerings, the result is just personalized in the sense that it is selected for me from centralized templates, and not personal in the sense of: unique and independent of the central templates.

BTW, the preference from the example — summaries concise and straightforward — is certainly easier to accommodate than understanding personal nuances, as I argued in my recent post about distilling.

Page with comments

I copied my recent PDF into a wordpress page ‘Opinions on AI‘ to enable further comments, and pointed to Stephen Downes’s valuable comments from a long Mastodon thread.

Wrong intelligence focus

There is a long paper by Carlotta Pavese (via OLDaily) talking about skills that are theoretical or intellectual vs. those that are practical and embodied. I agree that the priorities of education in developing intelligence are questionable, but I don’t think it is sufficient to just change the extension of the concept of intelligence. Rather, we should be aware that there are other mental qualities that don’t align with our Western modernity’s ideal.

Images: both from Wikimedia

The paper contains useful distinctions of various characteristics of what we call intelligence. Intelligent behavior is typically seen as goal-directed, flexible to circumstances, and adaptive according to previous experiences. Characteristics that seem to be especially intelligent, are thinking, executive functions, cognitive control, knowledge and cognitive architecture, and abstraction.

Several of these fit the idea that the human mind exists to manipulate the world rather than understand it to coexist with it: goal-direction, executive, control. Perhaps less obviously, the kind of thinking discussed in the paper is productive thinking, about how to do things, and has a product as a goal. And the generativity of thinking is based on its recursivity and its hierarchical structure — which entails collapsing and nesting.

Now of course I have to acknowledge that the traditional ideal of intelligence, with emphasis on theoretical and intellectual skills, does have a merit. A simple reason is that it helps being better prepared for an unknown future, and in particular, being able to cope with the invisible (future is invisible). But the mentioned attributes of ‘flexible’ to circumstances, and ‘adaptive’ according to previous experiences, do not address unknown circumstances and future experiences. And the mentioned ‘knowledge and cognitive architecture’ (mainly the assumed ‘declarative’ compartment of knowledge, which also entails more ‘sapere’ than ‘cognoscere’) does not promise long-term sustainable benefit.

So what remains, ‘abstraction’, does indeed play a very special role. For one thing, practicing to deal with abstract ideas also helps getting used to coping with the invisible. And abstraction may help to create indirect solutions that don’t solve a single concrete problem directly but many arbitrary problems indirectly.

A second benefit is that abstraction may help to transfer insights from one domain into another. But this effect is also achieved by a different kind of generalization: by metaphorical thinking. And while metaphors apply to two or multiple concrete situations, abstractions apply to none. And this is IMHO why abstractions are not everyone’s taste, and glorifying the abstract variety of thinking excludes many people.

So I think the focus on goal-directed, hierarchical, abstract thinking for manipulating and dominating the world, is missing out on a lot of valuable mental contributions of a wide range of cultures that are not trimmed to what Western competition calls ‘intelligent’, such as, like, indigenous ways of knowing, or care perspective.

The paper talks about ‘intelligence elitism’, and of course the common understanding of the term intelligence implies a scarcity of the individuals who possess a lot of it. Which, in turn, likely promotes unfairness simply by game-theoretical ‘power of the few’. But democracy is able to overturn this by the power of the many, and to decide that everyone shall be treated equal, no matter if they are ‘created’ equal. So I think it is not necessary to try to extend the concept of intelligence. Instead, awareness is needed of how limited this notion is, and that other mental qualities might be much more beneficial for an unknown sustainable future.

Distillator

One thing that I really hope AI will do for me, is sifting through large amounts of text. While I was just thinking about Stephen Downes’s post (distill, analyze, summarize), another post arrived (scanning, skimming, not reading, literature reviews).

It would already be a great help for me if the machine could just tell me what topics were covered in which text or video, ideally in which section, or identify some time stamp markers for the audio.

So for me, it is not too important whether AI would reliably interpret the texts and find out what is salient/ what ‘stands out’/ is noteworthy fo me, or ‘grok’ the ‘gist’ or whatever. I doubt that it would be able to simulate my interpretation, because it does not share my bias and emphasis.

I need to visualize that difference. In the following animation, the blue dots shall depict facts/ knowledge/ propositions from a domain or topic. The yellow, orange and red dots represent the linguistic ‘glue’ that turns the bare information into sentences, and this is colorful and shaped by diverse emphasis and biases. To my understanding, the machine model does not know or use such kinds of emphasis, but rather a version that is averaged from its giant training data points. So, not red and yellow, just orange.

Now if the input text was also generated by a machine, the distillation of a gist might work better. But sooner or later I would question why anyone still creates inflated text versions when a truthful, loss-free summary is possible — and probably the prompt for the generative product would suffice already.

(Similarly, if a picture is ‘worth a 1000 words’ but can be specified by the few words of a prompt, wouldn’t the prompt words suffice?)

Instead of creating inflated versions, only to distill and shrink them at the receiver’s side, the receiver might inflate the gist himself if he really desires that? OK if the energy consumption is high, it is better to do the transformation just once at the sender’s side. But OTOH, if personal weightings can be applied some day, it is again better to do it on the receiver’s side. And if one machine talks to another machine, they would probably avoid the inflated text stage, altogether?

My own experience with automatically inflated explanations, when I tried out GitHub Copilot’s code comments, were very frustrating. The copilot just paraphrased the obvious, while my own commenting tries to convey the salient, non-obvious gist to my later self.

So, I imagine a division of labor where the machine identifies the topics in a text, and the human interprets what was said about them. Roughly a distinction between theme and rheme. This division also aligns with my favorite rule of thumb: let the AI sort, not rank.

Duly Noted

The first good book about note-taking is out: Duly noted by Jorge Arango.

I was often frustrated with guidebooks that treated note-taking as a mere accessory for reading and writing, listening and talking.

In particular, the countless study guides annoyed me who preached that you just have to collect reading notes and then write the outline — without talking about what comes in between these two steps, obviously taking it for granted that it has to happen entirely within one’s skull, and there is no tool possible. Or the newer guides who harped on one special magic tool or technique for effortlessly becoming smarter.

This new book covers the entire process, from noticing and capturing to sharing and tending. And in particular, the center step of thinking and generating more ideas:

“Seeing notes on the board suggests other ideas. You move them around to form clusters, suggesting further ideas. A virtuous process follows.” (Kindle Locations 335-336).

The book thoroughly explains the differences of all the possible variations and types involved in the process. For example, items describe

• 9 x What Are Notes For?, 7 x Note-Taking Media, 7 x Types of Links, 3 x AI Assistants, even 4 x “Essential Distinctions” themselves,
• 5 x The Production Process, 6 x A Process for Thinking with Notes,
• 4 x Capture Use Cases, 3 x How Collaboration Changes Note-Taking
• 9 x Tactics for Generating Insights, 8 x Building Sustainable Habits, 9 x Selecting System Components, and 4 more lists of tips.

It is no incidence that the book is strong with clear distinctions: its author comes from Information Architecture.

The important distinction between ‘capturing’ and the later stages makes it clear that capturing needs to be quick and should not distract from thinking. And the discussion of ‘externalizing’ carries on Annie Murphy Paul’s idea of “seeing anew”.

Although my own method differs at times, the book is a wonderful resource for digging deeper.