Fast Learns, Slow Remembers: A Pace Layered Approach to Digital Preservation

Editor's Notes

• #2: Hello. I am going to talk about a thing called the Pace Layer Model and describe how it can apply to digital preservation.
• #3: Let me start with the obvious. Digital Preservation is a long term endeavour  The goal is: to ensure that records of the past are cared for in the present so that they are trusted in the future. This requires an uncompromising view of the long term.
• #4: Long term thinking is hard.   In the words of neuroscientist Daniel J. Levitin  “Our brains evolved to focus on one thing at a time.” And we have ingrained biases to decide which things deserve our attention. It is important to recognize that thinking about decades and centuries, not days and weeks requires real effort. We can’t do that all the time. There are some processes that unfold over long periods of time, and others that change much more rapidly. This is what the pace layer model tries to describe. 
• #5: The Pace Layer Model is a form of systems thinking. It is a way of modelling all of the components of a complex system. Instead of defining components based on, for example, their function or their physical location, it defines components based on their different change-rates and different scales of size. Each component is considered to operate at and change at a different pace. It considers the differently paced components to exist in layers.  Instead of breaking under stress like something brittle, these systems yield as if they were soft.  Some parts respond quickly to the shock, allowing slower parts to ignore the shock and maintain their steady duties of system continuity.
• #6: The first example of the Pace Layer concept came from the English architect Frank Duffy in the 1970’s. In the 90’s Stewart Brand, founder of the Long Now Foundation, expanded on Duffy’s work, in his book and BBC TV series ‘How Buildings Learn’.  Think of a building, like the one we are in.  You can model the components of this building into pace layers. The foundation of this building is over 100 years old and is not designed to change anytime soon  The services, like the lights and the WiFi, change much more frequently, perhaps over a few years. The chairs in this room will be rearranged in a matter of days. 
• #7: Stewart Brand also worked with Brian Eno to develop what they called the ‘healthy civilizations diagram’, which is a Pace Layer model with 6 layers.  They published this in a book called  The Clock of the Long Now: Time and Responsibility. Here, they are modelling the  components that make up a civilisation. Fashion and Commerce are where creativity and novelty live.  Infrastructure includes essential things like transportation and communication systems that are not justifiable in purely commercial terms. Notably this includes intellectual infrastructure like education and science. Which are High Yield, delayed payback activities. Governance serves what Stewart Brand called the larger, slower good Culture moves at the pace of language and religion. A ‘vast slow-motion dance keeps century and millennium time’. An example: the names of months (a concern of the governance layer) have varied radically since 1500, but the names of signs of the Zodiac (culture) are unchanged in millennia. Nature: droughts can last 300 years.  Carbon cycles are even slower. Through a series of chemical reactions and tectonic activity, carbon takes between 100-200 million years to move between rocks, soil, ocean, and atmosphere in the slow carbon cycle. The fast layers innovate; the slow layers stabilize. The whole combines learning with continuity. In a durable society, each level is allowed to operate at its own pace, safely sustained by the slower levels below and kept invigorated by the livelier levels above.
• #8: This example was created by Gartner Research. They promote what they call a Pace Layered Application Strategy.  In short: many large enterprises struggle with tension between business users of enterprise applications, who want to solve specific problems quickly,  and the IT professionals charged with providing these applications, who are trying to standardise on a limited set of comprehensive application suites in order to minimize integration issues, maximize security and reduce IT costs To quote Gartner:  Many organizations have used the single integrated suite as a means to ensure process and data integrity. Unfortunately, this is often the root cause of many organizations' inability to adapt their IT architecture to meet changing business requirements. The Enterprise, as a system, has components of different size and different rates of change. Gartner describes 3 layers. An Example Financial accounting is highly standardized, and is nearly always a system of record. Order entry often has a number of unique characteristics based on the company's products, channels and pricing, and would frequently be considered a system of differentiation. Collaborative demand planning is a relatively new application that is constantly evolving as companies look for new ways to use the Internet, smart devices and social networks to interact with customers and sense or shape demand. It is a perfect example of a system of innovation. In addition to segmenting applications into pace layers, Gartner’s pace layered application strategy also includes the ideas of connective technologies and governance as key elements. I won’t spend more time on the Gartner approach now, but these ideas - how to connect the layers, how to govern the changes within layers and across layers, this is the next topic I will discuss.
• #9: In a pace layer model, we have different components operating with different rates of change, meaning across different time scales,  and also operating at different scales of size.  How do the fit together as a single system?  What happens when one layer rubs up against another? How do changes in one layer affect the changes in the layers above or below? It is in the contradictions between these layers that civilization finds its surest health Stewart Brand There is some space between the layers. It is in those spaces that the magic really happens. The word liminal comes from the latin word for ‘threshold’.  It is the doorway or space between two places.   There is a paradox here. Liminal spaces can be both unsettling and transformative. They are a crucial part of system design. Being stuck ‘betwixt and between’ can be destabilizing and disorienting.  It can result in stress, dysfunction, at worst breakdown. When the liminality is recognized and respected, it can be a space for “construction and deconstruction, choice and transformation”.  The two most effective mechanisms for recognizing and respecting the space are Slippage and Feedback
• #10: There needs to be slippage and feedback between layers.  Layers need some independence. The tendency is for lower layers to control and constrain, and for faster layers to be more visible, they are a better match for our ingrained biases towards the immediate and the recent.  Each layer must respect the different pace of the others.  Slippage doesn’t mean disconnection.  There needs to be feedback between the layers as well. Feedback from lower/slower layers to higher/faster ones, to constrain ‘innovation’ so that it operates within the necessary context of the layers below.  Brand gives an example of this: Changes in technology have enabled commercial forces to act with more and more power, more effect.  We need laws to prevent one company from cutting down all the trees, or polluting all the water. 
• #11: Stewart Brand provided an example of what happens when fast layers dominate.   Imagine an architect who produces highly detailed designs, and is not involved in the actual construction process, and is not present when the building is occupied.  Falling Water, Frank Lloyd Wright's architectural masterpiece, is considered by some to be the "best all time work of an American architect." Yet the famed residence is so plagued by leaks that it is structurally damaged and, consequently, uninhabitable. Wright's roofs were known for their leaks. When asked once about this, Wright reportedly responded by saying, "That's how you know it's a roof!" You can think of this as the art/fashion layer dominating the slower layers of commerce or governance.  The Architect cares about the way the building looks, and does not care if it functions.
• #12: When the governance layer is not able to accept any feedback from higher layers, it stifles innovation.  New ideas need room to germinate and grow.  Rules exist for a reason, usually there is some past experience that the rule was created to prevent from happening again.  But without space to try new things, systems stagnate and collapse under their own weight.
• #13: I have two examples of applying the pace layer model to Digital Preservation.  Both include positive examples of slippage and feedback.  Archival Information Packages are a valuable strategy for aligning faster layers with slower ones.   The Fashion layer produces artworks.  Commerce produces records of transactions.   Infrastructure produces science and engineering records.  Governments, I have heard, produce records as well.  These outputs, either individually or collectively, are how the layers of governance and culture are shaped and changed.   As John Sheridan pointed out in his presentation this morning, records are produced by governments, on the time scale of commerce or infrastructure, records of transactions or events, for example.  And those same records become part of the governance process itself.   This process operates in a lower and slower layer.   By preserving the outputs of the faster layers we extract something that is culturally essential from the systems that operate at the level above By preserving those outputs in system independent and self describing packages, we allow for both slippage and feedback. Archival Information Packages enable slippage by decoupling content of enduring value from both the systems that produced it and the systems that preserved it.  Technological changes at the governance or infrastructure level do not affect the contents of an AIP.   AIP’s  also provide feedback. They constrain - by requiring the metadata necessary to understand their context to be preserved alongside the content.  These requirement bring the constraints of lower layers to the understanding of the outputs of the layers above.  At the level of commerce, we encourage innovation, within the legal constraints of, for example environmental protection laws.  By requiring AIPs to include metadata, the lower layers ensure that the technical, legal, and cultural context is communicated across time and space.
• #14: I have a final example from the early days of the agile software development movement, that is directly relevant to digital preservation efforts.  On this slide I have used the term Human Centred Design, which is a whole area apart from agile development, but it is the place where these two ideas intersect that I am most interested.  I am talking about collaborative and iterative systems design, which uses the techniques of prototyping to enable innovation and fast creativity, and expects and incorporates user feedback to adapt the designs and the system based on lived experience.   The quote on this slide is from Paul Dyson, one of the early adopters of the Extreme Programming movement in the 90’s. He wrote about his experience finding the pace layers in software delivery. He was involved in running an XP project in a small but conservative organisation, in 1997.   Their XP process included daily standups. The managers were invited.  They liked that they were able to tell what the system would do differently in two weeks, but the detail in the daily standups was not helpful to the managers.  The developers found it difficult to get what they needed out of the daily standups if they tried to meet the needs of the managers. The managers had customers they needed to deal with who didn’t care about the development process and just wanted some software to help solve some business problems.  They operated on a scale of weeks and months.  The development team was planning by the week, delivering new software daily., coding in pairs and refactoring and integrating hourly.  They found the best solution was to simply to recognise and respect the different rates of change in the organisation. They introduced some slippage between the two layers operating at different rates of change.  Managers stopped coming to the daily standup, and got a simple weekly status report.   They also added a ‘release planning’ practice, where the managers would give the dev team feedback in the form of high level statements about required features, and through discussion with the managers, the dev team would give very high level estimates of time and complexity.  It resulted in a very rough plan for what each customer release would contain, enough to communicate with customers about what was coming while also preserving the fast pace and agility of the development process. The faster layer got to deliver within the constraints of the slower business processes, and over time some of the advantages of the agile development process percolated down into those business processes.
• #15: The Pace Layer model is a useful tool for enabling long term thinking.  It provides a way of understanding how to design, build, and operate sustainable digital preservation systems, balancing innovation and change with stability and continuiity.  The Pace Layer model is a useful tool for enabling long term thinking.  It provides a way of understanding how to design, build, and operate sustainable digital preservation systems, balancing innovation and change with stability and continuiity.