market drive innovation management, from the inside in
DESCRIPTION
A presentation of a so-called paper, earlier today at APMS 2008, Helsinki, FinlandTRANSCRIPT
Market-driven innovation management,
from the inside in
Georgios G. Tziralis & Ilias P. Tatsiopoulos, NTUAAPMS 2008, Sep 15-17, Helsinki
Enterprises and organizations seem like market-agnostic, semi-permeable bubbles, in a market-
powered universe.
the problem
• Innovation management and idea selection
• in general:
- n potential evaluators of ideas
- m ideas,
- evaluated by i variables each
• optimal solution: the evaluation of an omniscient agent
background
• need for efficient information aggregation
• tools in use so far vary
- nominal groups, brainstorming, discussion boards and voting systems
- poor for information evaluation (static, low motive, etc)
• other options?
markets
• bring people together
• trigger conversations
• sum up information
• transmits it through prices
- idea selection is a distributed computation and markets seems like a fitting tool...
the approach
• each trader begins with her own information
• by observing market prices,
- she learns or infers what information other traders are likely to have
- she can then update her own info and beliefs
• over time, all information is revealed through trading, market price and traders beliefs converge
questions
1. Does a prediction market converge to a consensus equilibrium?
2. If yes, how fast is the convergence process?
3. What is the best possible equilibrium of a prediction market? and
4. Is a prediction market guaranteed to converge to the best possible equilibrium?
the model
• model design eternally balances on the trade-off of realistic detail and meaningful simplicity
- this one is biased towards simplicity
• information structure
• market mechanism
• agent strategies
information structure
• (boolean) state space of the world S = {0,1}m
• traders’ information space X = {0,1}n
• initially, each trader is privy to one bit of information xi (input bit)
• conditional distribution of information
- Q(x|s): {0,1}m x {0,1}n →[0,1]
- common knowledge to all traders
ehm?
• x1=1 (x1=0): product (not) viable for production
• assume one-dimensional state space (s=s1)
• P(x1=1 | s=1) = 0.9
• P(x1=0|s=1) = 0.1
- aggregation uncertainty!
• agent i starts with xi, but what about input bits of other traders?
market mechanism
• target: predict the value of f(s)
- f(s) : {0,1}m x {0,1}n →[0,1]
- s: state of the world
- f: boolean, common knowledge to all traders
market mechanism
• pari-mutuel market (Pennock 2004)
• two securities k, l, one for each binary outcome
• price(k), price(l) ∈ [0,1] (continuous variable)
• initially price(k) = price(l) = 0.5
• if finally f(s) = 1, all money are re-distributed to the owners of security k, and vice versa
market mechanism
• a multi-period Shapley-Shubik market game
• market proceed in synchronous rounds
• on each round, each agents buys one share of a security, either k or l
• after each round, new prices are announced
• price(k) = quantity(k) / (quantity(k)^2 + quantity(l)^2)
• process continues till equilibrium
agent strategies
• risk neutral
• myopic
• bid truthfully
• act rationally, each time submitting their valuation of securities’ expected payoff, updated via Bayes’ rule
• these are common knowledge
deductions
1. price convergence
2. convergence speed
3. best possible prediction
4. convergence to the best prediction or not
price convergence
• Without the arrival of new information, the market converges to equilibrium in finite steps
• At equilibrium, all traders have the same expectation about the value of f(s), which equals the equilibrium market price
convergence speed
• The market converges to equilibrium after at most n rounds of trading
- n is the number of traders
best possible prediction
• The best possible prediction that the market can make is the forecast at direct communication equilibrium
- namely the equilibrium that is reached immediately when all market traders directly reveal their private information to each other, i.e. E(f(s)|x)
convergence to best prediction
• An information market is not guaranteed to converge to direct communication equilibrium
discussion
• introduced a novel market approach to idea selection for innovation management
• formalized a simplified model of such a market process, with promising results
- convergence to equilibrium in relatively few steps, a state fully descriptive of available to traders information
• paved the way for further research (more advance model, simulations, experiments)