Saturday, January 23, 2010

Eight observations - 8th

The eighth observation of eight is about how computers and narrative connect. At the time I made the original eight observations I was working at IBM, so "how computing can help" was an important issue. Designing and writing software related to narrative is something I've done a lot of in the past several years, so I find I have a lot to say about these issues. (That is code for "this is going to be a long post, so stay with me.")

Grails and graffitis

My original observation was that there are two categories of goals related to computers and narrative: grails and graffitis. By "grail" I meant the holy kind, as in things we'd like to do someday but probably won't be able to do (at least not well) any time soon. The grails I named then were automatic story creation and automatic story understanding. At the time there were a few dozen people working on artificial intelligence systems to create and understand stories, and there still are. Erik Mueller at IBM Research has an excellent list of nearly 75 attempts to either create or dissect stories using computer algorithms. I haven't kept up with that part of the field, but as far as I know, for the most part these attempts are still in the theoretical realm and are not used (much?) in everyday life.

By naming the other class of computer-narrative connections "graffitis" I was referring to a story about the Graffiti handwriting recognition system for the Palm pilot, which was en vogue when I gave the talk. At that time Graffiti was a code word for a new idea in computing. The story was that after many partially successful attempts to recognize the insane variety of handwriting real people actually use, the creator of Graffiti had a brainstorm: ask people to learn a simplified shorthand-like system. This improved the accuracy of handwriting recognition tremendously. I myself used a Palm pilot at that time for taking notes during talks, and I found the Graffiti "gestures" easy to learn and even quite elegant. I could write nearly as fast with Graffiti as I could type (which is saying a lot).

I'm not sure what happened to Graffiti, but I found the idea inspiring. If you try to create a tool that meets people exactly where they are, you can get close, but making those last few steps to your goal may consume vast amounts of energy. However, if you can get people to just take a few steps to meet you, the tool can do much more to help them than it could otherwise. Call it satisficing for computers, or, computers help those who help themselves. Pretty much the same idea is what catapulted Google to fame.

So, my "graffiti" categories in that talk were two: tools that help people work with stories, and storytelling environments.

Updating the categories, or pitiful attempts at

I started working on this post by describing all of the computer-narrative interface projects I've worked on in the past ten years (it comes to around fifteen, depending on what constitutes a project). I had intended to summarize what I learned in each of the projects about what works for computers and narrative. However, the task turned out to be much more difficult than I expected, for two reasons.

First, all but one of the projects were done as a consultant. Consulting is a neither-here-nor-there kind of life: I can't speak for a client about its products, but neither have I the right to talk much about them on my own. Also, I haven't always given clients finished products: often I have contributed only designs or proof-of-concept prototypes. And even when I have completed software, clients have sometimes taken it and done who-knows-what with it afterward. So the list, as I tried to write it, was pockmarked in places where I couldn't quite say what I wanted to say about things I was not sure I could talk about.

The other difficulty in describing these projects was that the ideas I loved but could never interest clients in fully funding started to clamor for mention. In those parts of the post, unlike the pockmarked bits, the writing got too excited and went on too long. The whole thing was getting to be a mess, so I finally gave up the attempt.

However, the happy conclusion is that going through the exercise of describing these projects to myself did help me come to conclusions that I can talk about, hopefully in a way that is useful to you. (To make the clamoring ideas happy I plan to start another series of blog posts called "The Island of Misfit Story Ideas" and talk about them one by one.)

So in summary, as I look back on this observation nine years later, the grails are pretty much unchanged. That fact may reflect more of my not working in those areas than anything else, though I will have more to say about one of the grails later. The graffiti categories are also unchanged, but they have expanded into two sub-categories each. I'll go through those four categories one at a time and talk about generalities across software I've worked on. For each I say what I think is "best," meaning what I've seen work best.

Storytelling environments as narrative play arenas

By "narrative play arenas" I mean software environments in which people are supported as they grapple with issues by experiencing and manipulating stories and narrative elements. This can include looking at patterns in collected stories, comparing viewpoints, asking what-if questions by building models and other constructs, and building new stories.

This is serious play, because play is one of the most serious methods people have for making big decisions. It's a shame the word has such a dominant connotation of being for entertainment only, because it has the same serious importance for children too. That may be shifting: at this site consultants affiliated with the LEGO company actually offer workshops in serious play (yes, with Lego) for corporate groups. If the serious-play concept has gotten this far, perhaps I don't need to even say anything more about its validity.


Like the best toys, the best software for serious narrative play is expansive and generative. When people use it, new ideas should come flooding out. The ideas will not all be good ideas, but that stage comes later. And like the best toys, such software should be fluid, flexible and open-ended. An ineffective tool for narrative play is like one of those children's toys on which the child can do nothing but push buttons and listen (I admit buying a few of these when my son was an infant, when the insecurities were running high and I fell for the "genius" gambits). I've seen software that purports to support decision making that resembles such button-pushing toys. People go through the rigid processes as prescribed, but at the end their minds haven't grown a bit, and the ideas they have coming out are not much different than the ones they had going in.

The point of narrative play is not to come to a solution, but to change your brain. Coming into the use of good narrative play software should be like coming into a room with miles and miles of paper and millions of crayons of all colors, and thousands of sticks and blocks and balls and wheels and construction sets, all of them ready to combine and recombine, and plenty of room to build towers, destroy them, and build them all over again. As a perfectionist by nature, one of my mantras when drawing is, "There's always more paper." There should always be more paper in the world of narrative play.


Another essential quality of software for narrative play is discomfort. Again this is like the play of children, which to be effective has to include some non-parent-alleviated disappointment and cognitive dissonance. Good narrative play software should disrupt as much as it engages. Surprising and even upsetting juxtapositions should arise. Assumptions should be challenged. Mirrors, sometimes even distorting ones, should be encountered. This can be hard to pull off in software, because software has a hard time begging people to come back. The best way to enable disruption is to let the data do the talking and keep the software clear of any sermonizing. Instead of saying "You're looking at this all wrong" it is far better to simply throw up two graphs side-by-side (what they said, what you said) and step aside.


Finally, good narrative play includes synergistic collaborations. This is probably the hardest part of the experience to get right. An expert practitioner can help a roomful of people achieve breathtaking improvements in collective sensemaking. Capturing that magic in software is a much harder task, and I'm not sure if anyone has really done it yet. My experience has been that it's better to give people flexible tools that support synergism, but to recognize the limits of software to affect behavior and step aside and let people work our their own approaches to collaboration.

Of the narrative play environments I have worked on, the one that got closest to completion supported groups of government analysts whose task was to scan periodically collected stories and other data for upcoming problems in many spheres of national interest: security, transportation, disease, crime, and so on. The part of the system I worked on involved narrative play in that analysts would use collected "open source" items such as news stories to build constructed artifacts that represented current understandings or hypotheses, and look both for surprising differences in perspective and for outlying items that needed closer attention. This is play at its most serious.

Storytelling environments as narrative substrates

Narrative substrates are places where people share raw, spontaneous stories of personal experience as they happen or come to mind. For most of human history the only available narrative substrate has been face to face communication. In most of the world that is still true. Even when we routinely take "let's find out" to mean "let's type it into a search engine," sharing stories still happens more in person and over the phone than it does through software. However, people do tell stories using software, either unadorned in email or in discussion groups, or with "digital storytelling" software or through story sharing web sites.

The best metaphor for a software-based narrative substrate is a fertile soil, because supporting the optimum natural growth and reproduction of stories is similar to supporting the optimum natural growth and reproduction of plant life.


One requirement for a fertile narrative soil is a diversity of experience. Ranking, popularity, and performance destroy diversity through filtering stories so that only the best get told. These measures of social comparison are useful when people are making selections among products, services, groups, and sources of information. They are inimical to story sharing.

To illustrate this, please follow me as I construct an extended metaphor. It will be worth it, I promise! This metaphor follows the same line as the "stories are seeds" analogy in Working with Stories. Quoting from here:
In the soil, tiny charged particles called micelles usually have many areas of negative charge (called sites) on their surfaces. Positively charged ions (cations) are drawn to these negative charge sites and stick to the clay particles (are adsorbed).
Any community has many sites of attention through the minds of the people in it. Stories are drawn to these sites and are remembered.
In most soils, 99% of soil cations can be found attached to micelles (clay particles and organic matter) and 1% can be found in solution. Mineral cations in the soil (mainly Ca2+, Mg2+, K+ and Na+) maintain an equilibrium between adsorption to the negative sites and solution in the soil water. This equilibrium produces exchanges -- when one cation detaches from a site (leaving it free), another cation attaches to it. Therefore the negatively charged sites are called cation exchange sites.
The great bulk of stories are remembered, while only a small percentage are actively being told at any time. Communities maintain an equilibrium between remembering and telling stories. This equilibrium produces exchanges -- when one story is told, another can be remembered.
Because any cations loose in the soil solution are vulnerable to leaching as water flows out of the soil, a high cation exchange capacity is always desirable. Cation exchange sites act as a mineral buffer for the soil, storing minerals important to plant and animal growth for long periods of time.
Because stories in circulation are vulnerable to being forgotten, a high narrative exchange capacity is always desirable. Such a capacity acts as a narrative buffer for the society, keeping stories important to human life for long periods of time.
When ammonium nitrate fertilizers are added to the soil, the ammonium ions (NH4+) are strongly attracted to cation exchange sites because of their high valence (4). The ammonium ions displace many other cations which are then leached out of the soil and lost to plants.
Purposeful stories are strongly attracted to attention sites in the community because of their strong emotional impact, compelling structure and memorability. The purposeful stories displace many other stories which are then leached out of the community and lost to the people in it.

So, if you have been reading my rants against purposeful stories taking over our lives and have wondered why it matters, this is why it matters: purposeful stories reduce the diversity of naturally occurring stories. That in turn reduces the diversity of experience available to people trying to make sense of the world, make decisions, and get along with each other.


Plant roots were once seen as passive receivers of water and nutrients, but this is no longer the case, says Jorge Vivanco (among many others):
The rhizosphere is a dense and complex environment, in which plant roots negotiate a shifting sea of stimuli, including pathogenic and non-pathogenic microbes, competing plant roots, various invertebrates, and a wide variety of soil conditions.
Similarly, a community is a dense and complex environment in which stories negotiate a shifting sea of stimuli, including dangerous and non-dangerous rumors, competing stories, various attempts at control, and a wide variety of community conditions.

An important determinant of root health is soil texture, or the mix of particles of different sizes in the soil. A diversity of soil particle sizes is optimal, with "loam" describing a roughly equal mix of sand, silt and clay. What loam provides to plant roots is essentially a set of tools for deriving adequate water, air and nutrients from the soil. When only one particle size is available, something critical is missing: in sandy soils, water and nutrients are lost; in clay soils, air is lost.

Software that creates loam narrative soil has a diversity of tools people can use to tell stories, look at stories, watch over stories, remember stories, make sense of stories, and deal with problems in the story substrate. When the diversity of tools is limited, functions are lost. When stories are ephemeral, memory is limited; when stories are static, the air of reorganization and reuse is missing.


In gardening, microclimate is everything. The soil facing South next to your white garage may be in a different plant hardiness zone than the soil shaded by a large allopathic walnut tree in the East corner of your garden. One of the first things a new gardener learns is the folly of treating all locations equally. Farmers and gardeners engineer microclimates by adding windbreaks, flood channels, and cold frames.

In a narrative substrate, context is everything. A group telling stories about divorce in Kentucky may be in a different story hardiness zone than a group telling stories about a street in Calcutta. One of the first things a narrative substrate needs is the capacity to adapt to many contexts of storytelling. Creators of narrative substrates may want to engineer storytelling contexts by adding boundaries, privacy, signs of respect, and other context-determining measures. Thus any software that supports narrative substrates should maximize, without confusion, the ability of groups to make storytelling work within the unique contextual meaning of their group.

Narrative feature detection

When you want to find something out, it is almost too easy to collect or discover many hundreds or thousands of stories. Dealing with what has been collected, on the other hand, can be overwhelming. One antidote to narrative overload is narrative feature detection.

Feature detection in images, or computer vision, has been a goal of much research in the field of artificial intelligence, with mixed results. Some aspects of human vision have been unexpectedly difficult to duplicate. For example, looking into a moving, changing crowd and picking out a familiar face is something infants can do but computer algorithms are only starting to approach. However, other tasks have been easier for computers than for humans and thus useful to us. Detecting objects in fog or other low-contrast situations is an example that has practical applications.

Taking this as an analogy, tools that help people detect features of meaning and emotion or narrative elements in written texts (or transcripts of spoken texts) can help people see through the fog of narrative information to find looming obstacles (or opportunities). In general quite a few language and narrative tools can be helpful in this way. For example, say some thousands of stories are collected from web discussions or customer calls. Highlighting places where people made statements about values they held, or where people told stories, or where people talked about a person in a way that signaled the person was the antagonist in a story, would highlight features relevant to the needs of the moment.

Edge detection

In image processing, edge detection means simply highlighting contrasts, which are usually input into another process that tries to figure out why certain areas have higher contrast than others. In narrative work, edge detection has to do with finding contrasts in stories and story metadata: between positive and negative values, for example.

For this work the most useful tool is juxtaposition. In image processing an edge detector is essentially a small box (or "filter") that roams across the image, marking all pixels whose neighborhood shows high contrast. This produces the glowing-edge pictures you see in Photoshop.

 In narrative processing, tools can help people look across an "image" of story "pixels" for similar contrasts. The difference, of course, is that a narrative image reconvenes with every new look at it. Good narrative feature detection software makes it easy to create a range of narrative image assemblies and highlight the edges on them for consideration.

Blob detection

In image processing, blob detection is the isolation of areas that are minimally similar to other areas, such as high plateaus or valleys. The same can be done in narrative feature detection. As with edge detection, the landscape can be rearranged based on what elements matter.

I've written about the idea and practice of narrative landscapes for feature detection in this paper (see the section called "Mapping space"). It's a promising area that I think more people might want to make use of. Essentially, the idea is that if you ask questions about stories with gradients in two dimensions you care about, and then ask a question related to stability, you can create a complexity landscape of hills and valleys. Ridges and mountains indicate where people are telling stories of change or instability, and valleys or holes indicate where people are telling stories of inertia, hopelessness, or security. Looking at a landscape so produced can provide useful insights into what people believe, fear, and care about with respect to a topic you are exploring.

Of course, this form of mapping is only one of many such. The general idea is that you can use collected stories to ask questions about a variety of conditions you care about.

Pattern recognition

Pattern recognition in machine vision is about looking for known patterns on which action can be taken. Computers might be looking for particular parts to align for assembly, or faces to match with a database, or weather patterns that indicate a gathering storm. In some ways this is the most exciting part of computer vision, because people are finding out ways to enhance human pattern matching.

One of the more interesting ways pattern matching is being applied is in detecting patterns of behavior (of people, vehicles, computer programs) that indicate a situation requiring attention. People are designing systems that analyze patterns of head movement to wake up sleepy drivers, diagnose illnesses with recognizably unusual movement patterns, point out hospital patients who require immediate attention, and of course find people behaving suspiciously in airports.

Narrative pattern recognition has to do with looking for patterns about what people are saying in the stories and in answers to questions about stories. For example, in one project I can remember, stories marked as rumors featured strong links to dangerous outcomes in the use of a product, while first-hand stories showed no such links. This denoted a false rumor that could be countered by increasing the exposure of customers to true stories of the safe use of the product. In another project we found a striking set of differences in opinions about corporate responsibility and customer service that depended entirely on whether the respondent owned a home or rented one. This was an important societal distinction that determined how people felt they related to many agents in social life, including companies whose products and services they used.

I've seen hundreds of similar patterns, many of which repeat across projects, and I've built up a sort of library of expected and unexpected narrative patterns. Another common one often comes up in relation to age and participation in organizational life. People start out their careers with great energy and idealism, but little power. In the middle years their energy wanes while their power to affect change grows too slowly, causing frustration and burnout. Older people tend to bifurcate into two groups: those in power, thus satisfied (and sometimes blind to idealism), and those out of power, jaded, and unable to muster the energy to keep trying. Whenever storytellers answer an age question, I know to look for that pattern, and for any manifestations of distortions to it.

Detecting abnormal human behavior similarly relies on detecting an array of subtle cues in facial expression, body language, and word choice that indicate known patterns. This article describes what is being done in airports to detect the ways people behave when they have something to conceal. In The Wizards Project, Paul Ekman and Maureen O'Sullivan examined some twenty thousand people and found only fifty who could detect a lie with at least 80% accuracy. Ekman and O'Sullivan also found that while some people have a natural talent for reading the complex nuances of human microexpression and body language, others can learn to do this through training.

I've never taken any sort of evaluation on whether I have a special ability to pick up on deception, but I did once have a boss who insisted on taking me to every meeting because she said I could tell her when people were hiding something. The idea of "wizards" scares me because it seems to set up a superclass who might exert power based on invisible skills -- always a setup for corruption. But it does make me wonder if that old story is linked to the ease with which I pick up on patterns in bodies of collected stories (though it could just as easily be simple practice). I remember one conversation where I was telling others in my group how I used grounded theory on a collection of stories, and I said, "Next I circled all the phrases that jumped out," and one person stopped me and said, "Wait, don't you realize, nothing jumps out to us." I found this hard to believe, but it sounds weirdly reminiscent of what I've read about "naturals" in human lie detection. I've also read that facility in grounded theory, which similarly relies on picking up subtle cues of meaning in spoken or written text, requires a degree of natural talent that can be approximated by training.

I didn't bring up my possible natural talent in narrative pattern recognition for self-aggrandizement, but to make a point about designing software for narrative pattern detection. Expert pattern recognizers, whether natural or trained, need a flexible set of tools that respond to their good instincts with alacrity. But for novice pattern matchers flexibility can be damaging. They don't know where to start, and they can't call up tools that respond to their instincts, because they don't have any. They look at the mass of stories and nothing jumps out at them. I'd go so far as to say that software for novice and expert pattern matchers has contradictory specifications.

One software feature that can benefit both expert and novice pattern matchers is the embodiment of expertise in stored templates. For example, a software program could support the creation of search templates which experts create and novices apply. As novices become more familiar with the process and instincts begin to emerge, they can start creating their own templates. The story-scanning system I mentioned above (for government analysts) included aspects of template creation and sharing, in the form of conceptual models and other constructed artifacts, for this purpose.

Templates also have the benefit of breaking up some of the solidification that takes place as expertise builds, by juxtaposing templates drawn from different perspectives. For example, two expert analysts, one in health care and one in transportation, might create separate search templates, then use them to examine differences in pattern recognition among collected stories from multiple perspectives. Building adversarial templates from source documents, for example speeches by terrorist leaders, can also help to shake up overly ossified assumptions about why people do what they do.

Grounded story construction

The final group of software projects I've worked on (and this may surprise some) has to do with crafting purposeful stories. I actually think what is available to people who have a need to craft purposeful stories -- to teach, to persuade, to engage -- is far poorer than what could be created (and I will visit some of these ideas on the Island of Misfit Story Ideas later). However, as with narrative pattern detection, there is so much skill and natural talent involved that it is difficult to build tools that work equally well for everyone.

The metaphor (there must be a metaphor, you know) that springs up for story construction is the suite of tools digital artists use to create works of art -- photography, visual design, and so on. As above I have chosen the three aspects of such software that I think provide the greatest benefit.


My two favorite parts of Photoshop, which I use to "mess around" with photographs, are the Filter Gallery and the Color Variations screen. (In fact, this is the main reason I usually choose Photoshop over the Gimp, though I like that for some other things, like better extensibility.) Using these interfaces, I can very quickly play with alterations to my image, thus:

Juxtaposing these variations either in space or in time creates an expansiveness that enables play with the image I am creating. You may notice that this is similar to the play I talked about way up there in the section about narrative sensemaking. There's a reason for that. The best creation involves sensemaking, and the best sensemaking involves creation. They work together.

All of the exercises I know of related to making sense of stories (not least those described in Working with Stories) can result in material that works for purposeful story creation. Because the exercises are typically used for sensemaking only, the result is typically discarded; but in most of the sessions I've seen it has been clear that the end product could have been used to create polished purposeful stories. In the case of the composite story exercise, the result is literally a story, but an unpolished one.

Emergent constructs such as personifications, situations, motivations, values, and so on could be used to build stories. People in a sensemaking exercise could derive the constructs, then combine them into stories using such a simple device as writing the constructs on cards and making up some rules for combination. You might ask people to select two personification cards, one value card, one motivation card, one situation card, and so on, and build a story out of them. What you have at the end of the sensemaking session is a set of half-formed, but grounded, meaningful, resonant stories.

What happens next hinges again on the distinction between experts and novices. A person skilled at writing stories -- a short story writer, for example -- would be able to take the outcome of any narrative sensemaking exercise and "run with it" to build persuasive, compelling purposeful stories. Such stories would be far superior to any other crafted stories in achieving their goals, because they would be grounded in what matters and makes sense to people involved in the issue. Stories with excellent narrative form without relevant grounding are like movies with excellent special effects but plodding, predictable stories -- the surface shimmers, but the depth is featureless.


The other way to support grounded storymaking is to create tools that help non-experts craft compelling, purposeful stories based on the outcome of narrative sensemaking exercises. Again templates come into play, this time in the form of folktales or fables. The structure of fables is ancient and well-known, and using fable form is the best way I know of for expert storytellers (through the ages) to help novices craft well-formed stories. (I once worked on prototype software that helped people apply folk-tale templates to collected anecdotes for the purpose of quickly presenting complex understandings. I won't say more about it here, because it lives on the Island of Misfit Story Ideas and will have its own post later.)

In the visual arts, templates come into play in the cultural language of visual expression. Certain attributes of created images convey messages of purpose and context quickly and effortlessly. Consider what these devices convey:
  • sepia coloring - old times, history
  • "torn" edges - a photo album, "you are there" reporting
  • tilted and sometimes backwards letters - for or by kids
  • psychedelic colors - alternative perspectives, nonconformity
  • black and white - artsy, authentic
  • extreme close-ups - edgy, penetrating
  • "foggy" edges - cute, romantic
... and on and on through hundreds or even thousands of variations instantly recognized by people and used by design experts. Even sites like give you a quick way to send messages through your choice of style templates.

The universe of folk tale forms does the same thing with stories, though those styles are less well known. I'm not sure very many experts in folk tale form remain, since people have become unused to telling them. Embodying some of that knowledge in software can help people communicate in the same way that choosing "Artistic" and "Stylized" filters in Photoshop can help you embed a cultural message in an image.

Now here's something interesting. Just now, when I looked up "software for fiction writing" I found oodles of offerings: software to help you keep track of your characters, discover plot holes, record research, break writer's block, free up creativity, brainstorm, rearrange ideas, profile your story, and so on ad infinitum. But when I tried adding "folktales" or "fables" to the search text, the links I found were different: they all related to helping schoolchildren use fables in lesson plans about simple writing. I wonder if "serious" writers don't use folktale structures because they think such structures are too simple, or only for children, or not "real" storytelling. Most of the people who buy software that helps them write stories seem to be hoping to write the next great novel, which may explain the preference. That may be true, but for smaller efforts, templates based on folktale structure represent an untapped resource.


The third element of purposeful storytelling on which I have worked (at the prototype stage) is in the area of providing objective inspection functions.

Software for visual artists supports image inspection in two ways. First, most image software includes diagnostic aids such as histograms, edge detection and print previews. All of these draw the artist's attention to problem spots. In addition, many artists participate in mutual image reviewing in a group or community. Sites such as deviantART are essential resources for artists bent on refining their skills.

As with images, story inspection overlaps with feature detection. Applying filters that reveal edges, zones, and patterns in stories can help writers improve the coherence and effect of their stories. Generally, the analogue of image diagnosis in story writing has to do with emotion and value, since those are the colors of story palettes. I've worked on software that inspected stories and other texts for expressions of value, emotional intensity, and other aspects of meaning.

And as with images, software that helps people gather reactions to told stories can help them refine their skills and output. Most of the so-called story sharing sites on the internet are more about mutual inspection and review than they are about narrative substrate creation. On a mutual-review site, ratings, popularity, and "hot or not" comments help people hone their skills with needed feedback.

Revisiting the story understanding grail

Finally, I will return to one of the grails of my original observation: automated story understanding. I helped with one project related to this, though only tangentially. The project involved automating classification of stories, for example marking those with extreme expressions of emotion or values. The results were surprisingly good for issues like emotional intensity and positive-negative value; but still, the accuracy didn't exceed something like eighty percent.

What I saw was that the first 80% and the last 20% of automation can have contradictory results. The first 80% may be immensely helpful, but pursuing that last 20% can destroy the utility of the first part, and then some. For example, a mostly-automated indexer that suggests classifications could reduce the time needed to complete a process (say, triaging customer complaints or sonograms); but a completely-automated indexer that classifies items without human oversight could miss the one critical case out of a million that any human would know was in need of immediate attention. Missing that case could eradicate the gains created by all of the other automation.

So, this grail is still far off, in my opinion. Besides, I think it's a moving target. I would not be at all surprised if robots fifty years from now can understand all the nuances of human communication. But I would not be at all surprised if robots fifty years from now demand equal rights ... which would leave us pretty much back where we started. Said Gregory Bateson in Mind and Nature:
There was once a man who had a computer, and he asked it, "Do you compute that you will ever be able to think like a human being?" And after assorted grindings and beepings, a slip of paper came out of the computer that said, "That reminds me of a story ... "

In practice

As fascinating as this tour through computer narrative world may be, you are saying, how does it help me in my quest to work with stories? Well, that's hard to say. Of the projects I drew from here, only two are available for use: Cognitive Edge's SenseMaker Suite (though it has moved on since I was involved with it) and Rakontu. The former is useful for narrative play and feature detection. My dream is for Rakontu to encompass all of the graffitis I mention here, but that is far off; at the moment it mainly supports a narrative substrate, with a bit of narrative play and feature detection.

However, you don't necessarily need dedicated software to achieve these benefits. One of the reasons I listed features of useful software for each category is that you can gain those benefits by using software you already use in new ways, and you may not even need software in some cases. Word processors can be used for play; search engines can be used for feature detection; books of folktales can be used for story construction. You can build your own solutions for any of these goals, if you know what you need to build. Hopefully this journey through what I've learned will give you some resources to help you.

No comments: