ISIS KNOWLEDGE MANAGEMENT PRODUCTS

Where are the data we missed as we looked at surveys?
Where is the information we missed as we looked at data?
Where is the knowledge we missed as we looked at information?
Where is the wisdom we missed as we looked at knowledge?

This document is a primer to ISIS skills and tools in the fields of data collection, information and knowledge processing.. In its present form it may be considered as a tentative syllabus for a number of seminars. These are intended to make explicit and understandable the corpus of knowledge acquired and to a large extent built by ISIS. The material presented here is targeted at experts in allied fields with the scope in mind of defining and formulating possible co-operations in order to offer end customers a set of well proven approaches, tools and procedures to solve systemic problems.

Data collection and quality assessment; Data processing; Modelling proper; Building Scenarios:an Alternative to Modelling; Decision Support Systems; Wide Range Cultural Prescriptions - Reengineering a Positive Growth of the Web

1. Data collection and quality assessment

Current statistics (public, private, corporate) derive their data from official and private sources, research projects, EC surveys, international organisations and use methodologies which on occasion are not transparent. Validations, when attempted, often follow divergent, ill defined rules. Consequently shortcomings in data sets are frequent. This is true even when sources are reputable international outfits (e.g.: EITO, European Information Technology Observatory; EUROSTAT; OECD)
ISIS has developed procedures aimed at comparing and cross-footing data from different sources as well as mathematical tools apt to detect lack of quality and consistence in time series. This analysis can be attempted, e.g., by fitting Volterra Lotka equations to data (see Section 3.3 below)
When measured data are lacking, sometimes the attempt is made to derive information from interviews administered to a representative sample of a given population. Again pitfalls are frequent. ISIS statisticians have developed methods to assess reliability of samples and of the survey structure.
Once reliable, consistent data sets have been defined, indicators can be built to highlight cross impacts with other variables. Care has to be taken to avoid any attempt to use said indicators instead of missing data. There are no shortcuts to the problem of reconstructing a usable set of hard data, while original observations or measurements have been lost or, worse, generated in inconsistent meaningless ways

2. Data processing
ISIS has developed an original software tool - STATISTICA Cube - for data processing, interrogation and statistical analysis, purposely designed for reporting activities of organisations, public and private companies, statistical offices (local, national or international institutions). It allows to build up indicators and stratify data according to classifications of data which users can easily define by themselves and which are the most relevant for the problems and issue at hands. It includes user friendly functions to analyse uni-variate and multi-variate structure of data frequency, identify and isolate outliers, produce output charts and statistical tables with the most common statistical parameters, as well as, in the most advanced and complete version of the tool, to apply some of the most common techniques of multi-varied statistical analysis in an easy and practical way.
Thanks to its developing structure inside of Windows - by providing a wide range of icons and interactive menus - the software requires neither strong computer skills nor specific handbooks to be read. Besides, all of the calculation procedures can be automatically set, in order to create periodical reports (monthly, quarterly, yearly) without no strain. This automatic procedure guarantees reliability and quickness of statistical data output, as - after the generation of the a required set of procedure - these can be retrieved and applied on new set of basic data, while it is not necessary to create time after time new peculiar processing procedures, which may require technical skills and, whereby, a certain amount of time for their validation.

The software allows the user to design and implement on a standard PC a full statistical reporting system based on the data stored in the data models. The basic features are:

1. Handling of basic data included in the data models, which includes creation and updating from external sources of data description, qualitative attributes (stratifications) and quantitative attributes (variables), listing and printing of single records, modification of single data values.
2. Stratification of data based on selected variables.
3. Formulation of indicators and standard thresholds for statistical benchmarking and comparison of data. Standard thresholds can be simple averages or any other value of a single indicator, or a multiple set of thresholds according to a selected stratification of data.
4. Frequency analysis of statistical data, with the processing of frequency histograms and various statistical parameters (mean, mode, median, coefficient of variation, skewness coefficient) for any variable and/or indicator chosen from those included in the database.
5. Analysis and neutralisation of outliers for selected variables/indicators.
6. Selection, listing and ranking of data based on simple or multiple threshold values and the selection of a variable/indicator for data (increasing or decreasing) ordering.
7. Processing of statistical tables, which allow to build double-entry tables with:
- two selected stratifications respectively as rows and as columns of the table and one variable whose statistical parameters are computed and shown in each cell of the table , or
- one selected stratification (as rows or columns, as it is more convenient for practical printing purposes) and a set of variables whose parameters are computed and shown in each cell of the table.
Simple or stratified tables (triple-entry) can be elaborated.
8. Processing of tables of contingency (frequency tables): these are a standard tool of statistical exploratory analysis where the subdivisions of two stratifications are taken respectively to define the rows and the columns of a table, and the number of entities classified according to both stratifications are shown in the cells of the table (cross- or joint-classification). Absolute and relative numbers of entities (frequencies) in each cell can be considered. In particular, different types of frequencies are computed to show the percentages of entities in each cell on: i) row marginal total; ii) column marginal total; iii) total number of entities; iv) the product of the row and column marginal total. The latter give the so named "specialisation indices", sometime called independence/association indices or - when geographical subdivisions are concerned - location quotients, that measure the relative specialisation of each entity/location with respect to the attribute in question (e.g. prevalent economic activities)

Statistical tables and frequency tables are complemented by quick options to produce graphs from selected rows and/or columns of the tables, in order to highlight the most interesting trends and results. Then, the most interesting graphs as well as tabular data can be easily transferred to standard tools as EXCEL for further elaboration and refining, if needed. The software allows also to transform into EXCEL files data sets extracted directly from the database with the selection, listing and ranking option.
The application of the software was specifically required in the context of several contractual activities of ISIS for various national and international clients, including the Italian Ministry of Interior for statistics of local finance, CISPEL - the Italian Association of municipal services - for monitoring the quality of key municipal services such as water delivery, waste collection, public transport, energy, EUROSTAT for statistics on external trade, and many others.

3. Modelling proper
Modelling complex situations is a task which can be approached with a number of different tools. ISIS has been using econometric models, extrapolation procedures, correlation assessments, regression analysis, cluster analysis - well aware of the limitations of each approach and of the foresight necessary to avoid unwarranted conclusions.

3.1 System dynamics
System dynamics models were initially developed in the Sixties by Prof. Jay W. Forrester of MIT. This modeling technique aimed at describing and forecasting the long-term evolution of industrial organizations, urban structures and even world socio-economic conditions. Various hierarchies of structure are recognized:

- Closed boundary around the system
- Feedback loops as the basic elements within the boundary
- cumulative variables (levels) within the feedback loops influencing:
- Rate (flow) variables, influencing levels and closing feedback loops
- Goals
- Observed conditions.

The mutual dependence of variables is described by means of finite difference equations which are processed in a computer by software which determines next year's values based on this year's and on the dependence relationships suggested by past history. On the basis of recorded history, values are assigned to time delays and to multipliers affecting the cause-effect dependence of variables on each other.
Some variables are exogenous, representing scenario assumptions. Others are endogenous. The coefficients of the finite differences equations may be determined on the basis of theoretical knowledge of the processes. In many cases said coefficients are chosen empirically in such a way that produced forecasts are verified by subsequent system's behavior.
System dynamics models represent a very attractive tool to make explicit the involved operation of multiple feedback loops, which humans are hard put do assess by intuition. ISIS has developed criteria to avoid some current pitfalls in modelling (avoiding soft variables, naively accepting certain high-gain feedback loops, putting excessive reliance on empirical relationships). Validation of system dynamics models is a critical activity requiring practical experience and recourse to sophisticated mathematical tools..

3.2 I/O matrixes
Input-output models (developed by W. Leontief at Harvard) are schematic quantitative representations of national economies. They are based on data produced by the System of National Accounts which record the annual GDP and all the magnitudes that contribute to form the wealth of nations - with the same standard United Nations accounting methodology applied all over the world. The schematic representations consist in large square matrixes. Both column and rows headings are some tens (and up to a few hundred) sectors of activity (agricultural, industrial, services, commercial) of a national economy. The matrix elements are transfers of wealth from the sector indicated in the row to the sector heading the column. The method is designed to portray in detail the actual inter-industry relationships of a real economy.
The rationale of the analysis is based on a simple accounting identity: for each sector or industry the sum of all outputs (sold to other sectors) must equal the sum of all inputs (purchased from other sectors), provided we do not omit any transactions. This is equivalent to the accounting identity for an individual firm, which states that total receipts must equal total costs plus profit. A crucial assumption is introduced to simplify the relationships between the sectors: all production processes have fixed technical coefficients.
In any application of the input-output method industries or sectors must be divided into two groups, one called the final demand sector, the other the processing sector or structural matrix. The division is intended to reflect a distinction between those outside (exogenous) sectors in which the level of activity is autonomously determined and those inside (endogenous) sectors in which the level of activity can be explained by the model. The industries within the structural matrix are regarded as a set of processors whose output goes to satisfy the requirements of the final demand sector.

Input-Output models can be adapted to analyse:

- specific sectors of the economy, as for instance the transport sector, building transport satellite accounts which offer a greater detail of freight and passenger transport activities and the wealth generated by them within the economy, while the other industries and service sectors are traced at a more aggregate level, summing up several rows and columns of the standard national Input-Output accounts. ISIS has studied and proposed the systematic realization of transport satellite accounts for the EU member states in the context of UNITE, an European Research Project currently (year 2002), providing an unified methodology to account internal and external (i.e. impacts on environment and health) costs of transport activities;
- regional and urban economies, including also wider flows of distribution of wealth from the production sector back to the people and the various social groups which earn their income directly through their labor inputs to the production sector or indirectly via government taxation and provision of public services and other benefits to the population. This approach entail to analyze with more detail the urban/regional economic base and social impacts, and it usually implies severe data problems. In fact, it is difficult to gather the numbers required for an input-output table for a nation as a whole (national input-output table are usually make available from statistical office with a time lag of at least 5 years). It is even more difficult for smaller regions, since in general the smaller the area, the less statistical detail is available in published sources. But new systematic methods of data collection, quality control and data processing, coupled with the potentiality of ICT technologies to provide new real time procedures of data collection, make today more realistic these kinds of application than ever before. ISIS, in the context of the European Research Project ACT-VILL provided a pilot example of Social Accounting Matrix (an Input-Output model extended to cover distribution of wealth within society) for the Rome metropolitan area.

3.3 Logistic Substitution Models
Logistic substitution models consider separately each variable and determine for each the most probable equation governing the development process. The model is agile, can be applied rapidly and easily expanded. It applies Volterra's equations, derived to analyse variations of biological populations and represented by S shaped logistic curves. They depict accurately the mutual influences of species competing for food in the same habitat. These equations describe accurately also growth and decline of populations of human products and artefacts. Based on time series available, the model determines the equation of the process and computes then future developments. If we call x the number of units belonging to a population, the equation is:

dx/dt = k x (N - x)

where the derivative is made with respect to time t. The solution is:

x = N/(1 + exp(A t + B))

where N is the asymptote or final constant measure of the population.
The ISIS LOGI5000 software used fits an equation to a time series and measures the standard error of data with respect to the equation. If the standard error is more than 1E-02, the process can hardly be described in this fashion. Volterra equations were used with success to describe development of a variable to fill an available ecological niche, but in some cases the fit is not satisfactory due to the presence of noise in the data.
The software used - and the mathematics on which it is founded - permit also to give a judgement on the quality of the data used. Consequently the model provides also an indication of the credibility of the forecasts and projections it produces.

4.Building Scenarios: an Alternative to Modelling
When dealing with very large systems there is no rational procedure to analyze quantitatively what would happen in any one critical situation out of the many billion possible ones, which cannot even be listed and which could not have been foreseen by systems designers at the time(s) the initial projects were developed. Large systems tend to proliferate in random fashion, as designers and decision makers separated in time and space think up and implement independently portions of the systems, additions, innovations, retrofittings. In this situation modeling is all but impossible. The only extant option is to build scenarios. Here we present guidelines for building scenarios and list some examples.

4.1 Distinguishability of scenarios
Scenarios analyzed must set out assumptions about the future that make them recognizable AND interesting. A scenario is distinctive, if it assumes the insurgence of a step function: a sudden phenomenon with marked consequences cascading from a given sector into other sectors. Cases in point may be: the steep increases of the price of gold and of crude oil in the early Seventies, wars, revolutions and so on. The genesis of step functions may not be very relevant to the outcome, while only the dramatic end results have to be interpreted and discussed.
For example, if we describe a situation in which GNP is halved over a 5 years period, we are presenting a deep recession scenario (similar to events which took place in 1929-1933). Discussing the causes and the consequences of this single event can be quite instructive. Instead a precise description of a set of assumptions that differ slightly from previous situations or trends, do not provide significant new insights.
For example: A nation's scenario in which - 10 years in the future - GNP has risen 12%, the inflation rate has risen 30%, unemployment decreased 3%, the M1 Money Supply increased 8% is hard to distinguish from a scenario in which the corresponding percentages are: 10% (instead of 12), 27% (instead of 30), unemployment decreased 2.5 %, M1 money supply increased 10 %. We may deduce from these lists of assumptions detailed socio-economic consequences, but we will not understand what big factors may loom in the future of the nation considered.

4.2 Non gratuitousness
This guideline states that new scenario developments have to be chosen among those which can be reasonably expected to take place. They must not be grabbed out of thin air. We should concentrate our efforts on analyzing new events which reasonable and learned thinkers consider as likely and relevant. We should not spend time to derive the ultimate consequences of a new - ideology being adopted by large masses of people, as this type of process would develop only over the long term.

4.3 Incomplete exogenousness
Interesting scenarios admit of a possible individual (or more often) preventive or remedial action of humans. So, e.g., we should be well advised to spend time in analyzing the effects of even major cataclysms, but not the effect of an irresistible force (like a comet's impact) - unless we can define areas and possibilities of human action to exert an influence over the development of the considered scenario.

4.4 Adequate analysis of main scenario feature(s)
For a scenario to be useful, it needs to be covered by a text explaining the rationale for the imagined sequence of events and to specify assumptions made also concerning cause-effect relationships and explanations of the main consequences of assumed events. It is not enough to define it simply by means of a title or a too brief description.
A full employment scenario should explain what decisions, interventions, implementations are posited and explain why these will result in full employment. We should not accept, instead, any scenario based on wishful thinking, which, for example, tries to deduct naively what would be the economic and social consequences of full employment.
An innovative energy strategy scenario could assume, for example, that cold fusion is feasible, but certainly not that it would instantly conquer the market. Rather it should deduce from the feasibility of cold fusion the steps and processes leading to the acquisition of a growing market share.

4.5 Pedigree - or similarity to well known processes in the past
Based on the tenet "Nothing is new under the sun" we expect new processes of variation to be similar to past ones. This does not mean that we should look for strict historic courses and recourses, but rather that well established, previously successful explanations should be given more credit than those based on postulated mechanisms.
A good case in point is represented by energy sources substitution patterns. These are well known to develop both in the growth and in the decline phase following logistic curves with fairly uniform time constants. The very constancy of these laws characterizes a credible pedigree of the quantitative analysis in that it has been successful in a large number of cases over a period of time of about two centuries.
A more complex analysis is needed when we try to anticipate wars and revolutions. The only sensible way to do this involves enrolling the help of those professional historians who are able to collect the relevant facts and to interpret them in significant ways (Prof. Carlo Cipolla may be taken as a role model of this ilk).
Very obviously we have to deal here with debatable issues and methods. It is easy to utter words of caution against ideological approaches. We must also remember that one man's tool of the trade may appear to another man as an idee fixe and to a third man as sheer folly. This is a normal situation whenever highly complex and debatable questions are on the carpet.

4.6 A List of Typical Scenarios
We list here a number of positive and negative scenarios featuring technological innovations as well as socio-economic or political occurrences. The list represents only typical examples, which were chosen, however, among a number of cases on which ISIS has carried out original work.

4.6.1 Large scale adoption of electric cars
In the automotive sector it is reasonable to forecast a substitution process leading to a fast development of electric cars. The precedent of ten US States where laws have been passed to enforce certain lower thresholds of zero emission cars by the year 2005, is pointing that way.
Another important factor that could foster the growth of electric cars is the probable increasing recourse to hydroelectric energy. World-wide about 90% of this source is yet to be exploited, while transmission of electricity with very high voltage direct current lines is now feasible over ever increasing long distances (up to many thousand kilometers).

4.6.2 Large scale diffusion of telematic networks
Society is just beginning to feel the impact of telematic networks: Local and Wide Area Networks, BBS's, Internet, WWW. This wave of innovation is impacting work habits, human relations, management and government rules and practice. The number of people who telework or telecommute is increasing fast. The gist activities of advertisers, consultants, designers, advisors and experts of all kinds are changing from day to day.
Here, then, it is not so much a question of formulating a scenario, but of recognising a vast process continuously gathering momentum. The forecaster should, then, tackle the more modest job of anticipating events and innovations which are already spreading, being designed. Government and societal decision making processes will be deeply affected in ways difficult to foresee now. Forecasters should try and participate in the redesign process - after. having investigated available evidence or surveyed plans and policies. ISIS has been active in the field from its inception about one decade ago: at present ISIS is engaged in a study contract for the European Commission: SEAMATE on Socio-Economic Impacts of ICT.

4.6.3 Innovative Government industrial policy for new enterprises
To make assumptions on GDP levels is unavoidable. However a special scenario is needed to depict the consequences of an innovative governmental industrial policy for new enterprises. This could take the form of interventions of the type planned and co-ordinated by the Japanese MITI (Ministry for Industry and International Trade).
The essential novelty of the scenario would be two-fold. First: the industrial policy would raise the professional and cultural level of the workforce and, hence, of the population at large. Second: the diffusion of technological innovation would increase the possible tools and prescriptions available in the overall portfolio of socio-economic and industrial solutions.

4.6.4 Innovative Government and private education policies
We have to face the harsh reality that education policies have been seldom formulated. When they were attempted, this was done in a patchy way with no balance between the importance of different sectors, nor between the relevance of subject matters to societal ends.
Of course, blueprints for cultural renewal have to start with schools - at all levels. In this field total quality programs have hardly ever been attempted and benchmarking has not been a standard policy requirement. Consequently excellence has been achieved in a number of well known (and famous) cases, but it has not been replicated on a large scale.
The use of mass media for education has been attempted in many cases with excellent results. Suffice it to quote the British Open University, the educational channels in the US, the public educational TV channel in Japan. This scenario prescription would call for:
- total quality management introduced in all schools
- founding new excellent institutes for advanced studies and research
- establish mass media cultural programs (all levels, all targets)
- redesign careers to include retraining and new professional life after retirement
- train high level scientists, professionals, experts and practitioners to communicate and spread their knowledge directly and through the media
- plan investments in all sectors (economy, industry, administration) to obtain results through education/knowledge diffusion rather than through direct short term action programs

5. Decision Support Systems
Data collection, data processing, proper modelling activities, scenario building methods developed at ISIS all concur in some cases to realise full Decision Support Systems (DDS) for specific problem areas, developing original software tools and data bases.
Two examples of DDS's, developed by ISIS and currently used by European Commission Directorate for Energy and Transport policies, are a tool to evaluate policies of rational use of energy in the household, industry, tertiary and transport sectors (MURE) and another software tool to assess real costs - internal and external - of intermodal freight transport and compare them with the all-road alternative on selected European corridors (RECORDIT).

5.1 MURE (Mesures d'Utilisation Rationnelle de l'Energie)
MURE is a policy support tool developed by ISIS on behalf of the EC to provide input to the formulation of national and European policies in the area of Rational Use of Energy (RUE). It includes a comprehensive and continuously updated database on measures (i.e. legislative, regulatory, technological, financial, etc.) adopted by Member States to promote energy savings and environment friendly initiatives. It covers all end-use sectors (residential, transport, industry and tertiary). MURE further includes a simulation model allowing to build RUE scenarios whereby the potential impact of policy packages is evaluated both in physical (energy demand) and monetary terms, based on assumptions on the energy performance of technologies and their market penetration rates. Policy scenarios are compared to the BAU (Business As Usual) case. Scenarios produce forecasts at 2020 of energy savings, pollutant emission reductions and the private costs associated to the implementation of the corresponding policy packages
MURE is extensively used by the Commission for the ex-ante identification of RUE priorities as well as for the ex-post evaluation of policy packages, in particular through backcasting mechanisms. The EC has adopted MURE as its primary policy tool for RUE evaluation and for dissemination of RUE best practices. Policy scenarios have been run in many areas, e.g. to assess the impact on energy demand of the building codes, the potential penetration of Best Available Technologies in the industrial sector, in the health sector, etc.
Results have provided direct input to a number of official policy positions of the EC, including the recent Green Paper on the Security of Energy Supply, the Action Plan on Rational Use of Energy, and the so-called Building Directive.

5.2 RECORDIT (Real Cost Reduction of Door-to-door Intermodal Transport)
RECORDIT is a Decision Support System developed by ISIS in the framework of an EC funded research project. It is based on a detailed accounting framework allowing to calculate the real costs of intermodal transport along European corridors. Costs include both internal costs for the production of the transport service (personnel, administrative, fuel and other consumption, use of terminals, insurance, etc.) and external costs generated by the movement of goods (air pollution, noise, accidents, congestion, global warming, etc.). The DSS allows to simulate the impact of policies and actions on such costs, and therefore on the competitivity of the intermodal operators, especially when compared to the all-road solutions, which currently prevail in the European market. The RECORDIT accounting framework is highly detailed, thereby allowing to build policy scenarios of all kinds, based e.g. on the optimisation of load factors, on the interoperability of railway networks across Europe, on the introduction of new technological standards for vehicles etc.
For the calculation of external costs, RECORDIT has designed and implemented an innovative procedure, which allows to produce estimates of these values with a limited set of basic input data.
RECORDIT is currently being used by the EC to prepare the forthcoming Framework Directive on transport pricing, based on the systematic and consistent application of the so-called "users pay principle", and for which a reliable estimate of the real costs of transport is therefore essential.

6 Wide Range Cultural Prescriptions - Reengineering a Positive Growth of the Web
The very fast growth of all E-activities (E-all) prompts attempts to: record, analyse, reason out, forecast. The goal is: planning best use of Web and its impacts on all sectors and also reaping advantages as a consequence of increasing demand unleashed by a widespread diffusion of ICT literacy and higher level professional performances. A more rational growth of E-all should be based on a more positive planning of Web growth. This should not rely just on technological progress, which is continuous and hardly predictable, but on establishing structures, relationships, rewards, protocols, milestones - organised in positive self-replicating patterns.
A qualitative preliminary description may feature:
- cell like structure (reminiscent of biology and of Quality Circles)
- communication channels to explain/motivate
- rewards for success in preferred directions
- re-engineering of economic relationships
- theoretical elaboration.

6.1 Levels of social ICT learning
Society needs ICT learning at many levels: from the basics of end users buying and getting information, to merchants offering wares and services, organisations re-engineering logistics, experts planning, designing, implementing, running networks. Growth of low levels creates a ground from which high performers proceed to high competence levels.
A first step is: define functional profiles, contents, task, curricula. ICT literacy is not a single proficiency, but a range of culture and skills. The ECDL (European Computer Driving Licence), a standard of knowledge/skills accepted in all of Europe, is taught and learnt with profit. More elaborate skills are required for high level jobs.

6.2 Interactive quality control of E-learning
To achieve success, the quality of teaching and learning has to be controlled interactively. Reprogramming schools and creating new efficient ones is expensive. E-learning boosts efficiency and minimises cost. We describe here an organisation of interactive distributed tutoring. A path is proposed leading to competence and good jobs even at low initial levels. Reaching higher levels entails personal growth and prestige and also exploits newly acquired skills to offer tuition to others. The process of knowledge diffusion is controlled on line by the original teachers.

6.3 An incentive-motivated ladder structure of learning/teaching
Motivation towards learning is essential. These enabling factors are vital:
- in the beginning tuition is not free (to avoid playing down value)
- guarantee high quality teaching by fame of sponsors and teachers and by the offer of paid jobs in subsequent stages
- boost status by offering prestigious diplomas, the quality of which will be increasingly recognised by government, companies and the public.

Five levels are identified:

6.4 Sponsors
The endeavour described above is a prerequisite to the continued success of E-commerce and of the E-economy in general - hence to high level jobs creation. It will be partly self-sustaining, but it would be reasonable for local and central governments to participate with subsidies.
Other obvious sponsors are:
- producers of H/W and S/W
- companies active in the ICT sector
- producers of durable goods, retailers, service companies aiming at growing revenues from E-commerce
- Internet providers
- media