Information- -Velocity Metric Velocity Metric Information-Velocity - PowerPoint PPT Presentation

Information- -Velocity Metric Velocity Metric Information-Velocity Metric Information for the Flow of Information for the Flow of Information for the Flow of Information through an Organization: through an Organization: through an Organization: Application to Decision Support Application to Decision Support Application to Decision Support Jeff Waters, Ritesh Patel, James Eitelberg, and Marion Ceruti, Ph.D. 14th ICCRTS, Washington, D.C. 15-17 June, 2009 SSC San Diego … on Point and at the Center of C4ISR SSC PACIFIC … on Point and at the Center of C4ISR

Presentation Topic Outline Presentation Topic Outline  Information flow  Information velocity, v(info)  Relationship between information and power  Reducing uncertainty in decision making  Can we measure v(info)? Yes and No  Information-flow model for decisions support  Can we measure factors the influence v(info)? Yes - Direct measures - Causal measures - Effects measures 2 SSC PACIFIC …on Point and at the Center of C4ISR

Information Flow Information Flow • = Information flow, summarized as the p difference between in conditional entropy, H( h | l ) , of variable, h , before the process started given the variable l 1 and after the process finished, given the variable l 2 . • p = H( h | l 1 ) - H( h | l 2 ) ) is high because many alternative COAs • H( h | l 1 are consistent with sparse data. is low because the few alternative COAs • H( h | l 2 ) are consistent with the new data set. corresponds to the reduction in uncertainty that • p results from the receipt of new data. depends on the specific task. • p has no explicit time dependence. • p 3 SSC PACIFIC …on Point and at the Center of C4ISR

Information Velocity Information Velocity v (info) is defined as the speed and direction of • information flow, p . • First time derivative of the information flow. • Explicitly a vector quantity. • v (info) = dp / d t • = d H( h | l 1 ) / dt – d H( h | l 2 ) / dt • For example h = COA, l 2 = a new data set. • Depends on the task tractability, T y , & the power of information to reduce uncertainty. • An important topic of research aimed at reducing uncertainty in time-constrained decision making scenarios as fast as possible. 4 SSC PACIFIC …on Point and at the Center of C4ISR

Taxonomy of Information Velocity Taxonomy of Information Velocity Information Velocity Information Flow: Velocity Defined by entropy & uncertainty reduction. No direct, general metric Direction: Magnitude or calculation available. Information should flow in the right direction to the right Indirect methods: Quantity: How person or network Assumptions site where the much information information is Approximations can move? required for tasking. Modeling & Simulation Bayesian Networks Time management: How Current paper fast can information move? Future research 5 SSC PACIFIC …on Point and at the Center of C4ISR

How Is Information Related to Power? How Is Information Related to Power? • 2 nd law of thermodynamics, d S = d U rev / T = entropy, U = reversable heat, T = temperature – S analog:  H( h | l ) =  W / T i • Infodynamic W = work, T Y = task tractability, assumed constant at given entropy –  W = F d X = J ( d 2 X/ d t 2 ) d X • = distance, J = information (analog of mass). – X • d H( h | l ) / d t = (J / T i ) ( d 2 X/ d t 2 ) d X/ d t • Combining these equations with p yields v(info) = d/d t{(½ J X d 2 ) 1 /T Y1 - ( ½ J X d 2 ) 2 /T Y2 } where X d = d X/ d t Power = d/d t ( ½ J X d 2 ) • Energy = ½ J X d 2 • Take away from the derivation: The rate at which information travels is proportionate to power. 6 SSC PACIFIC …on Point and at the Center of C4ISR

Can We Measure Information Velocity? Can We Measure Information Velocity? • Yes. In modeling-and-simulation experiments. – Where we can define and control all the variables. • No. Even with many assumptions, v(info) is too hard to evaluate in practice because: – The nature of data interactions is not always known. – Pedigree metadata elements may not be available. – The data sets may be incomplete. – Data and pedigree metadata are time dependent. – Data distributions may not be Gaussian. – The form of H( h | l 2 ) may be unknown. – Time constraints preclude detailed enough data analyses in command centers. 7 SSC PACIFIC …on Point and at the Center of C4ISR

Can We Measure Factors that Influence Can We Measure Factors that Influence Information Velocity? Information Velocity? • Yes. Focus on practical time management in command centers using the following metrics: – Direct measures – Causal measures – Effects measures • Consider a decision-making process model of time management in organizations in general with applications to command centers • This model includes: – Measures of time spent on various tasks – Visibility of information – Empowerment of people – Direct, causal, effects metrics as unitless indices 8 SSC PACIFIC …on Point and at the Center of C4ISR

Information Flow Model for Decision Support Information Flow Model for Decision Support 9 SSC PACIFIC …on Point and at the Center of C4ISR

Information Flow Model: Information Flow Model: Expanded Decision Substates Substates Expanded Decision 10 10 SSC PACIFIC …on Point and at the Center of C4ISR

Assumptions in the Model & Metrics Assumptions in the Model & Metrics Metrics for time management in decision making depend on the following assumptions. • Better time management increases v(info). • States and substates described in Figures 1 and 2 are independent of each other. • Time spent in green states increases v(info). • Time spent in red states does not increase v(info). • Decision makers have the information they need in green states. • Decision makers can define start and end times for entry and exit of various states. 11 11 SSC PACIFIC …on Point and at the Center of C4ISR

Direct Measures Direct Measures • IVMDirect = Tg / (Tg + Tr) = time spent in green states • Tg – Considered productive activities = time spent in red states • Tr – Considered unproductive activities • Advantages of IVMDirect : – Time is an important factor in velocity. – Simplicity – Does not depend on causes and effects. • Disadvantages of IVMDirect : – State boundaries are not always clear. – Red states may contribute to uncertainty reduction. – All necessary info may not be available in green states. 12 12 SSC PACIFIC …on Point and at the Center of C4ISR

Causal Measures Causal Measures Min (Vi, Vy, Ep) • IVMCausal = Max (Hh, Pcr, Bc) = visibility of information and decisions • Vi = visibility of decision maker and metadata • Vy = empowerment of people to increase v(info) • Ep = amount of human-to-human comms • Hh = level of pressure, personal & cultural risk • Pcr influencing the decision maker = level of barriers to rapid, concise, honest • Bc communication • Numerator: select smallest value of Vi, Vy, Ep • Denominator: select largest value of Hh, Pcr, Bc 13 13 SSC PACIFIC …on Point and at the Center of C4ISR

Effects Measures Effects Measures • IVMEffects = < N D > < Q D > < S D > > = Average number of decisions per unit time • < N D > = Average estimated quality of decisions • <Q D > = Average level of satisfaction of the • < S D individual with the rate of uncertainty reduction 14 14 SSC PACIFIC …on Point and at the Center of C4ISR

General IVM Metric General IVM Metric IVM% = { (Q3% + Q4% + Q5%) / 3 - Q1% - Q2% + 100 } / 2 What percentage of your day do you spend in meetings, reading • Q1 and writing e-mail, talking on the telephone, in teleconferences, and in other forms of conversation and communication with others? What percentage of your day do you spend preparing products • Q2 intended for sharing information, eg. preparing briefing slides, reports, agendas, minutes, and completing forms and logs? • Q3 What percentage of what people are doing in your organization is important and relevant to you and your assigned tasks? What percentage of what others decide is important across your • Q4 organization or enterprise? What percentage of what others decide is visible and easy for you to understand on a daily basis? (Use a single percentage.) What percentage of what you decide is important is visible and • Q5 appreciated across your organization or enterprise daily? 15 15 SSC PACIFIC …on Point and at the Center of C4ISR

IVM Measures Are Unitless Unitless Estimates IVM Measures Are Estimates Min (Vi, Vy, Ep) • IVMCausal = Max (Hh, Pcr, Bc) • IVMDirect = Tg / (Tg + Tr) • IVMEffects = < N D > < Q D > < S D > • IVM% = { (Q3% + Q4% + Q5%) / 3 - Q1% - Q2% + 100 } / 2 • All variables in IVMCausal can be estimated on a scale of 1 to 10 • IVMDirect is a ratio of times so units cancel out . • < N D > = integer • <Q D > & < S D > - estimated on a scale of 1-10 • IVM% depends only on percentages. 16 16 SSC PACIFIC …on Point and at the Center of C4ISR