A HISTORY OF RECORDS CLASSIFICATION

In order to understand where we are now in records management land, we need to know a little about the history of our noble profession, which is one of the oldest. Records pre-date writing! The first records were clay statuettes representing real assets such as cows and sheep, or fields of grain; these tiny models were enclosed in a clay parcel and sealed, representing a single transaction - not all that different conceptually from what we do today. When clay tablets were first used, different transactions were recorded by function on tablets of particular sizes, and each tablet was marked with a classification code, representing another attribute. Thousands of tablets marked with the identical classification codes have been located, and soon after this first use of classification, early repositories come about, with the records of transactions kept in purpose-built halls, separately by function. Essentially this model has continued for 5000 years!

The Internet has surrounded us with trillions of records and billions of links and professionally at least, we are flabbergasted. In the nearly two decades since the World Wide Web was launched by Tim Berners-Lee, we have failed to incorporate any aspects of the Web into records management. Or can anyone suggest the ways we have done so?

No wait, I'm wrong! As brochure sites and as a research tool, we excel. The United Kingdom's National Archives site, and most especially the "Services for Professionals" section http://www.nationalarchives.gov.uk/services/?source=services is a wonder, and of course we have the at times magnificent records management listservs such as those I am aware of in the UK, the USA and Australia. And being information professionals, the Internet's power as a research tool is well understood.

Our brother and sister archivists, along with librarians, museum professionals and other curators, have done an excellent job of turning the Internet into a presentation, preservation and access tool. The pictures catalogues at the state and national libraries in Australia are a joy to behold http://www.sl.nsw.gov.au/ .

However in our professional practice we are confused. We are all struggling with electronic records, struggling to make sense of and manage them, and struggling to regain control over the records spread throughout our organisations. And the point of greatest struggle - the lever to lift the records management world - is the classification system we adopt.

EARLY RECORD KEEPING

Clay tokens, stored in round clay “envelopes,” represented the objects they resembled.

"This page is an amalgam of two of Luciana Duranti's articles and is not my own work" - the author, David Povey.

Order itself is something divine ... the soul of archives ... is nothing else than order ... first it is proper to divide up locations, then affairs, and finally times. ... Then let us prepare indices and syllabi, let us make up lists and catalogues in alphabetical order. Adapting to each set of materials its own indices, whatsoever will be needed we will have before our eyes immediately. Baldassarre Bonifacio (1)

A primary function of all records systems is command and control, not merely documentation. As the anthropologist LeviStrauss suggests:

… consider the first uses to which writing was put: it was used for inventories, catalogues, censuses, laws and instructions; in all instances, the aim was to keep a check on material possessions (2)

In fact, if it is true that all ancient societies created writings which mirrored the two fundamental sectors of any public or private activity, operational and housekeeping, it is also true that everywhere the housekeeping records preceded the operational ones.

The first records were created to take note of how much was due and how much had to be received, how many persons, workers or slaves were assigned to various works, how many goods were stored in the warehouses, how many animals, houses, lands were owned, and so on. They were records with an internal character, which remained in original form with their creator, while other parties could obtain an abstract of them. (3) Progressively, external writings began to join the housekeeping records: contracts, reports of external functionaries, and letters.

In the third millennium two separate records offices were established for the operational and housekeeping records, and sometimes the office for the operational records was constituted by as many offices as the competencies of the administration (i.e.: internal affairs, foreign affairs, etc.). Each records office maintained the material created and received according to a distinct records keeping system, and preserved it in separate storage areas. This division is particularly evident in the Royal Palace of Ugarit, in Syria, where six records repositories were found, containing material resulting from six different functions: 1) administrative records related to the cities and the lands of the kingdom; 2) legal and financial records related to the capital and its immediate surroundings; 3) judicial and "notarial" records related to public and private affairs of the King and to the properties of all the Kingdom; 4) records in foreign languages related to the relationship with Hittites; 5) records related to the relationship with Palestine, Egypt, Cyprus and Mycenae; and 6) records created by the various records offices to be controlled and delivered (a kind of mail room). The records were preserved in groups intellectually corresponding to modern dossiers, one for each transaction. (4)

TOKEN SYSTEM

The clay tokens, typically 1-3 cm. in size, are from the earliest periods revealed by excavation (ca. 8000 B.C.E.) and coincide with the period of emergence of organized food production and animal husbandry in south-western Asia. Bureaucratic and trading activities necessitated the use of some system of records for accounting and tracking commodities, increasingly involving credit transactions, and for some 5,000 years clay tokens served this purpose. Schmandt-Besserat found among 8,162 tokens a total of sixteen categories of tokens in a surprising variety of types and geometric shapes. Some of them are spheres, discs, cones, cylinders, tetrahedrons, ovoids, and quadrangles. (5) Over time, there is a movement away from the use of pictographic symbols-like animals, tools, and people-toward increasingly abstract signs.

The dimensions and shapes of tokens are quite diverse. They come in sizes which might be roughly categorized as "large" (3-5 cm.) and "small" (1-3 cm.), reflecting differences in quantity. Factional markings (e.g., halves, three-quarters) are sometimes incised on the tokens' surface as are lines, notches, punches, etc. Near the end of the first developmental period, the tokens begin to change from a simple or plain appearance alone (for crops and livestock) to include those of a more complex, or more data-rich, quality (for processed goods, such as perfume, jewellery, garments). There were new shapes (e.g., biconoids, triangles, rhomboids, parabolae) and additional markings (e.g., parallel lines, stars, crosshatching). These changes foreshadow the later emergence of the more abstract and numerous signs of clay-tablet logographic writing (one symbol for each word). (6) These later changes also coincide with advances in urbanization, an emphasis on skilled craftsmanship, a growing population, and a rising number and complexity of commercial transactions.

These tokens were not used as currency or for counting and so should not be confused with small stones (Latin calculi) used by the Romans in calculations or with the beads of an abacus used for the same purpose. Rather, the individual tokens correspond to our "data," which, when meaningfully combined and structured, represent the "information" embedded in records or documents. Schmandt-Besserat is certain that the tokens are clearly part of "the earliest system of signs used for transmitting information."(7) How, then, were tokens used for recordkeeping?

MAKING RECORDS

In their different shapes and markings, the various tokens represented many traded commodities and craft goods. These include sheep, goats, wheat, barley, clothing, beer, wool, metals, rope, oil, bread, perfume, mats, furniture, tools, woven mats, and pottery vessels. In the beginning, each discrete token represented "one" of whatever commodity or resource the token represented; e.g., one jar-of-oil token = one jar of oil; two jar-of-oil tokens = two jars of oil. This is an example of the more primitive "concrete" form of numeration, which was eventually replaced by a more abstract numeration wherein some conventional symbol, rather than a like number of objects, represented "three" or "four" or "ten" and so forth. Either the size of the token (large or small) or markings incised on the token represented the quantity involved.

The fundamental purpose of token-based records appears to be the same as all other records: to separate what is known from the knower, to objectify it onto a recording medium, to communicate it to others, to retain what is known on a basis more permanent than mere speech, and to avoid the confusion of multiple languages and dialects. The earliest token system, for example, allowed trader X to record for future reference that he bought, say, five bushels of wheat from merchant Y. This transaction was recorded by stringing together five spheres-each sphere denoting one bushel of wheat-and sealing the ends of the string with a clay tag (bulla) upon which the seller's seal is impressed. The buyer, or his surrogates, could later claim the grain by offering as a record the string of spheres to the merchant or his agents. When the transaction was completed to the satisfaction of all, the tokens were disposed of. (8)

REFERENCES

1. Lester Born, "Baldassarre Bonifacio," 233-236. in Lucia Duranti The Odyssey of Records Managers Part II: From the Middle Ages to Modern Times Source: Records Management Quarterly Publication Date: 01-OCT-89

2. Georges Charbonnier, Conversations with Claude Levi-Strauss: (London: Cape Editions, 1973), p. 30; quoted in Denise Schmandt-Besserat, How Writing Came About (Austin, TX: University of Texas Press, 1996), p. 55.

3. Luciana Duranti. "The Odyssey of Records Managers," Part I: From the Dawn of Civilization to the Fall of the Roman Empire, Records Management Quarterly, 23, iii (July 1989), 3-11 and Part II: "The Odyssey of Records Managers: From the Middle Ages to Modern Times," Records Management Quarterly, (October 1989), 3-11.

4. Duranti "The Odyssey of Records Managers," Part I (above)

5. Schmandt-Besserat, "The Origins of Writing," pp. 5-6 and Schmandt-Besserat, How Writing Came About, p. 125.

6. Schmandt-Besserat, "The Origins of Writing," pp. 5-6 and Schmandt-Besserat, How Writing Came About, p. 125.

7. Schmandt-Besserat, How Writing Came About, pp. 15-16; Robert K. Logan, The Alphabet Effect: The Impact of the Phonetic Alphabet on the Development of Western Civilization (New York: William Morrow, 1986), 20-21.

8. Schmandt-Besserat, How Writing Came About, p. 8.

BIBLIOGRAPHY AND ADDITIONAL RESOURCES FOR STUDY

Duranti, Luciana. "The Odyssey of Records Managers," Part I: From the Dawn of Civilization to the Fall of the Roman Empire, Records Management Quarterly, 23, iii (July 1989), 3-11 and Part II: "The Odyssey of Records Managers: From the Middle Ages to Modern Times," Records Management Quarterly, 23, iv (October 1989), 3-11.

Green, M.W. "The Construction and Implementation of the Cuneiform Writing System," Visible Language, 15, iv (Autumn 1981), 345-372.

Harris, Michael H. and Stanley Hannah, "Why Do We Study the History of Libraries? A Meditation on the Perils of Ahistoricism in the Information Era," Library and Information Science Research, 14 (1992), 123-130.

Hussein, Mohamed A. Origins of the Book: Egypt's Contribution to the Development of the Book from Papyrus to Codex. Greenwich, CN: New York Graphic Society, 1972.

Lewis, Naphtali. Papyrus in Classical Antiquity. Oxford: Clarendon Press, 1974. Logan, Robert K. The Alphabet Effect: The

Impact of the Phonetic Alphabet on the Development of Western Civilization. New York: William Morrow, 1986.

Pemberton, J. Michael. "High (Professional) Anxiety? Image and Status in Records Management," Records Management Quarterly, 30, i (January 1996), 66-73; 80. ___ "Professionals and Clerks: One Happy Family?" Records Management Quarterly, 28, ii (April 1994), 56-59; 60-61. _ __. Pemberton, "Does Records Management Have a Future?" Records Management Quarterly, 25, i (1991), 38-41, 45. Pettinanto, Giovanni. Ebla: A New Look at History. Baltimore: The Johns Hopkins Press, 1991.

Posner, Ernst. Archives in the Ancient World. Cambridge, MA: Harvard University, 1972.

Thomas, Rosalind. Oral Tradition and Written Record in Classical Athens. Cambridge: Cambridge University Press, 1989.

Weitemeyer, Mogens. "Archive and Library Techniques in Ancient Mesopotamia," Libri, 6, iii (1956), 217-238.

Bibliography for "earliest records systems: A journey in professional history"
Pemberton, J Michael "earliest records systems: A journey in professional history". ARMA Records Management Quarterly. Apr 1998. FindArticles.com. 11 Jan. 2008. http://findarticles.com/p/articles/mi_qa3691/is_199804/ai_n8805113

Ontologies: how actions are described using words

An ontology is all the words used to describe something; in the case of records management, all the words used in the business classification process and records system, together with other sources make up an organisation's ontology.

A definition of ontology is "the specification of one's conceptualization of a knowledge domain." [quote from http://www.noisebetweenstations.com/personal/essays/metadata_glossary/metadata_glossary.html ].

However, while an ontology has all words available, not all make the cut, as ontologies control vocabulary and through this the representation of knowledge. Hence the word "specification" in the quote above - in reality, an ontology is a limited set of available words and relationships, because it would otherwise be too complex. This is the "fatal flaw" of ontologies: whereas a searcher may look for "car registrations by vehicle type" cross-referenced to "traffic accidents at locations on the mid-North Coast", the ontology design would normally not permit North Coast towns being in the set of vehicle types: an ontology is designed for the Browse function, while today's users have experienced the more powerful function of the "Google" Search, that the relationships between information are so inherently complex that they cannot be predetermined and that it is preferable to allow clustering of results after search to clustering in any other way.

We have had long experience with ontologies and only a short time to experience clustering of information via searching. Our ability to use ontologies is thousands of years old - ancient Bradshaw drawings in north-west Western Australia may be 40,000 years old, the cave drawings at Lascaux perhaps 30,000 years - and these drawings represent complex relationships between the artist and the depiction. Now we use a catalogue or file index to do much the same thing, representing relationships using language rather than clay models or drawings, and the language used in that process makes up the ontology. It is full of "this is like that" and "this is part of that" relationships, and at the simplest level that we are all familiar with, is a file folder filled with papers, and an index card with that folder's name and location inscribed upon it; the next level up a little more multi-dimensional, such as computer network drive name, windows file share name, directory name, folder name and document title. There are further "part-of" relationships that documents are part of - transactions, reference, background and others; so in turn are these activities part of larger functions in the organisation. And so on, the whole language we use to describe these complex relationships is sorted and filtered into a structure and this is the ontology. As information managers we inherit long experience with this work.

The purpose of an ontology is its use as a tool to negotiate the pathways of our knowledge about corporate actions which have been described when classifying and indexing documents about those activities, and we have computer systems able to search all our documents for the words we have used to describe those actions, and to then display the results. Definitely a 2-way tool, for referencing documents and then for searching for and finding them.

As a tool, an ontology is subject to design rules: a hammer with pointy ends is a poor tool for hammering nails. The ontology's design should facilitate access to information about actions, for then it is a good, useful tool. The basic rule is the Rule of 3, which is that no more than three steps are required to categorise, classify and title any record, and the obverse of that, that finding any document should never exceed more than 3 retrieval steps. The most effecient ontology for records management purposes has 3 levels.

One of the great questions facing information managers today revolves around the ontology and its purpose - Google appears to show that we can effectively search and retrieve information without an ontology - or if we use an ontology it is of the broadest kind, using words and concepts at the very top of the classification tree, and then only to reduce the set of records from which our searched-for document will be found. This has significant consequences on the architecture of our information systems.

Classification and control

The anthropologist Levi-Strauss wrote: "… consider the first uses to which writing was put: it was used for inventories, catalogues, censuses, laws and instructions; in all instances, the aim was to keep a check on material possessions". In other words, writing was primarily for the purpose of control.

How does that control work in a business classification scheme? We records managers claim that first and second level terms in the BCS are there to make records easier to find - I won't provide quotes to prove that assertion, because every reference to the BCS terms makes it clear that first and second level terms play a findability role: they are in the BCS to make information easier to find. But it is clearly difficult - sometimes impossible! - to find unique information using top-level terms from a BCS. They only work for information retrieval when they are clustered with unique identifiers - and in every case, the unique identifiers find the information without the BCS terms!

Let's look at what happens if we search for records using first and second level terms - we find that the terms are so often and widely applied, that the number of "hits" of high-level terms precludes their use as finding aids. The significant role they play is as control words, words to which, for example, the retention function is attached.

The first and second level terms when applied to records are there for our use: these terms tell us information about the record. By "us" I refer to records managers - not records creators. Yes, it is possible to argue that records creators have as high a stake in the management of the record as records managers do but the fact is, most of the information they touch is not a record - see the Australian Public Service Commission paper on this topic, there is a link at the top this blog page or go to www.apsc.gov.au/mac/noteforfile.htm .

And so users object to having to "recordify" everything that passes their scrutiny as time wasting and unnecessary red tape. This objection is made clear in staff surveys of the Australian Public Service - referenced on another page of this blog - and in the post-implementation review white paper from the Michigan Records Management department EDRM at www.michigan.gov/documents/ hal_mhc_rm_finaleval_72433_7.pdf .

Having users recordify their information is placing the burden on the user and the benefit with us. To which the user rightly, silently or vocally, objects. Records management has always been about control: Control was the basis of the registry system adopted during WW2 and refined in the 1950's, that is the immediate predecessor of our own records systems. I speak of the time when TRIM was a physical file management system and yet that information environment was multi-dimensional; it was widely recognised then that records have an informational purpose along with their evidential purpose. Many of us still operate as physical file managers while we develop the means of managing electronic records and it is proving a struggle to mix these two activities. The briefing paper from the Australian Public Service Commission called "Note for File" addresses in some detail in its 70 pages, the notion of what public officers are expected to do with records, and makes it clear that matters of some importance to the commonwealth of Australia should be kept but that many other matters do not need to be retained for any significant period of time and so do not properly come within the purview of the records manager.

What are we to do? Let's put the cart after the horse!

This final blog in the series devoted to aspects of records classification has a practical bent, especially for those records practitioners in New South Wales Australia who are moving from a paper-based to an electronic records keeping environment. In his excellent paper "Ontologies are Overrated" [http://www.shirky.com/writings/ontology_overrated.html ], Clay Shirky asks "when does ontological classification work well?", finding that "Ontological classification works well in some places, of course. You need a hierarchy to manage a file system."

Let's accept we need ontologies to manage our file system - in practical terms, we need a BCS. The primary purpose of the BCS is to enable management functionality about records, mainly the retention periods. And that's where the heading for this posting comes from - while retention is an important part of records management, the "continuum" philosophy here in Australia has led us to sentencing records on creation, and in the EDRM environment that has led us to imbuing first and second and even third level terms in the BCS, with retention mandates and has vastly over-categorised our record keeping. In a medium sized NSW government authority, a retention schedule could have more than 1000 entries, and each of these entries is tied to a multi-level classification term used in titling. No wonder our users are in revolt!

Now the question becomes "to what level of granularity should the BCS go?" - should the BCS have first, second, third and fourth level terms, and to what extent are we mandating it? And how will that fit with retention scheduling?

My answer is to insist on the first level term and everything beyond that is optional - provided, and this is a huge proviso - that users are encouraged to title their records adequately. The rule for users then becomes - "adequately title your records and assign them to a top-level term" - we records managers will then "harvest" additional metadata from the Office system the user is working in.

So I am recommending that for most documents added by users to the corporate EDRM system, only the title and first-level classification terms are the responsibility of the user. As users are experienced in titling their own records, it is only adding the top-level classification term that is a record keeping activity and this will minimise the overheads for users, and hopefully encourage their use of the corporate EDRM for record keeping.

From our records management viewpoint, where does this leave us? Have we now lost control of records added to the EDRM? Given the very complicated nature of our disposal schedules, it presents as a struggle and so we now have to really bite the bullet and get to work on the disposal schedules - if we have 17 top-level terms, through which we intend managing records, then those terms have to be rich in the management functions they provide us. I suggest that most of our disposals could be managed in this way: that records about assets should be kept until the asset is disposed of; that personnel records may need to be kept for long periods of time; that most financial records need to be kept for seven years or less; that most informational records are kept while they are useful; that policy records should be current and prior versions retained for long periods of time, and so on. There are broad categories of records and broad retention classes to which most records can be put and that we should standardise and simplify retention to meet these criteria; in practice, users will be able to over-ride retention periods, by declaring a record to be a special case suitable for retention as a record of continuing long-term value or a permanent archive record.