Project outline

Final report

Press coverage

related research




Agent-Based Simulation to Aid Policy Making in Areas Related to Free Software

Alastair Burt, ASWAD Project/DFKI, Germany

A free software model of development entails the sharing of knowledge in the form of source code, that is it depends on the workings of an information commons. Not everyone is convinced that all code should be developed in this fashion and, even worse, there are several trends related to patent and copyright law that will make if hard, if not impossible, for the commons to keep functioning in its current form. Agent-based simulation tools could be a valuable counterweight here in two ways. First, they could provide a good framework for developing more accurate models of the effects of various policy decisions. Second, at a more practical level, they could provide a means to visualise the workings of the commons. This is necessary because one reason why the commons is in danger is that the harmful effects of policy decisions are not immediately perceptible. It is hard to visualise innovation that does not take place.

The Free Software Commons
The essence of the free software model that differentiates it so strongly from the closed software counterpart is that information, mainly in the form of program code, is continously shared. To use a term currently made popular by Lessig [Les01] and Bollier [Bol02] it is based around an information commons. The sharing of information is one of the principle motivations of developers in choosing the free software model. The developers typically want to have as many people as possible inspecting the code they put into the commons, improving it, and developing other software that builds upon it. Indeed, the most popular license for free software developers, the GNU General Public License, is designed specifically to keep as much information as possible within the commons.

Policies that Endanger the Commons
Although free software is currently enjoying a lot of popularity among developers and end users, there are several trends in the policy making of governments that could prove very harmful to the free software model. Most of these trends derive from an attempt to build a knowledge-based economy on the propertising and privatising of the information commons [Les01,Bol02]. The threats come in three main areas. First, the application of patent law is being extended in many ways into the software field. Patents are actually designed to increase the information commons, but, in so far as empirical studies have been carried out in the software sector, they would indicate that patents have the reverse effect [BM99]. Regardless of the effect of patents on the software sector as a whole, they have a deleterious effect on free software development, unless the patent holder is prepared to allow the unrestricted use of the patented method in free software. Second, copyright laws have a big effect on software development. The current application of copyright has lead to the prevailing closed source development model, although that need not be the case [Les01]. Moreover, the recent legislation that prohibits the reverse engineering of software that processes copyrightable material, and proposed legislation that would give the music and film industry the right to vet computer systems could drive free software from large swathes of the computing landscape. Third, laws associated with warranties for software can also affect free software developers. To make authors liable for all code they contribute to the commons would obviously decrease the amount of code released to the community. Policies that keep authorship and liability separate are to be preferred.

Agent-Based Social Simulation
Agent-based social simulation [GT99] lies at the intersection of social science, agent-based computing and computer simulation, where the agents in question are software entities with autonomous control, acting in a goal-directed and / or utility-maximising fashion. Agent-based social simulation offers a third way to consider policy questions [Mos99]. The other two more traditional approaches to policy making are simplistic mathematical models from economics and informal qualitative models from social science.

There are good reasons for thinking that agent-based simulation could be more useful than the other two approaches as a tool to demonstrate the value of an information commons and show the way policy decisions affect it. First, agent-based systems are good at modelling non-linear systems. Many issues related to group behaviour within the commons have a non-linear nature; the technology lock-in derived from the network effect is one. Second, agent systems are a good way to combine several mathematical models for agent behaviour and the behaviour of the environment. Third, agent-based simulation, in contrast to models from economics, can handle the more natural case where each agent possesses a different decision-making procedure. Fourth, agents are the appropriate tool for examining the rationality of an agent, in particular its resource limited nature. Last, agent-based social simulation is a good way to visualise the behaviour of a complex system [EA96].

Addressing the issue of information sharing policies and the cooperative development of code is an interesting topic for agent researchers. There has been a lot of work recently that has looked at the internal resource allocation policies of agents, in particular how to adapt the notion of rationality to an agent that has only limited resources to work out a good plan of action for the future [RW91,SBB97]. information sharing policies and the structure of the information commons address the same issue at the societal level: how to organise the cooperative behaviour of a society of agents so that the bounded rationality of the individual agents leads to behaviour that is optimal for the agent society as a whole.

Potential Areas for Research
Which are the areas related to the information commons where agent-based social simulation could guide policy making more effectively? The following is a tentative list:
  • Modelling the effects of patents on the problem solving abilities of individual agents. This should take into account issues such as the costs of policing patents, how easy it is to obtain patents (the novelty criterion), the problem domain (is there sequential innovation that needs many patentable inputs), and the length of patents. Eventually it should answer the crucial questions: patents may be good, but does one patent policy fit all industrial sectors, all the time, in all countries? Do patents increase or decrease the information commons in the software sector?
  • Modelling lock-in effects associated with a particular technological sector and the tipping points in a particular domain.
  • Modelling the effects of copyright - are restrictions on reverse engineering justified in terms of cultural variety?
  • Modelling motivational factors in free software development and the reduced transaction costs associated with peer to peer production.

J. Bessen and E. Maksin.
Sequential innovation, patents and imitation.
http://www.researchoninnovation.org/patent.pdf, 1999.
Revised 2002.
D. Bollier.
Silent Theft: The Private Plunder of Our Common Wealth.
Routledge, 2002.
J. M. Epstein and R. Axtell.
Growing Artificial Societies: Social Science from the Bottom Up.
MIT Press, 1996.
G. N. Gilbert and K. G. Troitzsch.
Simulation for the Social Scientist.
Open University Press, 1999.
L. Lessig.
The Future of Ideas.
Random Houseq, 2001.
S. Moss.
Relevance, realism and rigour: A third way for social and economic research, 1999.
Inaugural lecture as Professor of Social Simulation.
S. J. Russell and E. H. Wefald.
Do the Right Thing : Studies in Limited Rationality.
MIT Press, 1991.
J. H. Siekmann, A. D. Burt, H.-J. Bürckert, K. Fischer, P. Funk, C. Gerber, C. J. Jung, J. Lind, and G. Vierke.
Comma-cogs: A cognitive architecture for social agents.
Technical report, DFKI, 1997.