Actions

Difference between revisions of "Machine Learning, Complexity and Market Behavior"

From Santa Fe Institute Events Wiki

Line 19: Line 19:
 
New York, New York
 
New York, New York
 
----
 
----
What impact will the increasing use of artificial intelligence (AI), machine learning (ML), and other computational tools have on financial markets? Specifically, how might these tools shape market behavior, and even the nature of the markets themselves? We are still crafting this meeting’s agenda, but existing work by a number of SFI community members provide useful insights on these questions. For example [https://www.santafe.edu/people/profile/andrew-lo Andrew Lo’s] recent [http://alo.mit.edu/wp-content/uploads/2017/06/Moores-Law-Vs.-Murphys-Law-in-the-Financial-System-Whos-Winning.pdf paper] on Moore’s vs. Murphy’s Laws offers a broad overview of the complexities governing the impact of new financial technologies on market behavior. <br/><br/>Most experts agree that the increased use of these computational tools will increase heterogeneity, and computational models are particularly useful. [https://www.sciencedirect.com/science/article/pii/S0167268112000546 This paper] by [http://www.brandeis.edu/facultyguide/person.html?emplid=3863836c60fea8a993359c6d2f71be423bc77a23 Blake LeBaron] on the application of heterogeneity of learning to modeling asset price dynamics is a good example of this approach. Less well understood, is the how the use of these tools will impact the collective intelligence of financial institutions and the collective computation of the markets they animate. [https://www.theatlantic.com/science/archive/2017/07/collective-computation/533169/?utm_source=atltw This overview] from The Atlantic provides a good summary of these emerging fields, and [http://advances.sciencemag.org/content/4/1/e1603311 this paper] by [http://www.terpconnect.umd.edu/~ebrush/ Eleanor Brush], [https://www.santafe.edu/people/profile/david-krakauer David Krakauer], and [https://www.santafe.edu/people/profile/jessica-flack Jessica Flack] offers a useful two stage model for thinking about collective computation.
+
'''Morning Sessions - Collective Intelligence'''<br/>
<br/>
+
The new field of collective intelligence (see ''The Atlantic’s'' [https://www.theatlantic.com/science/archive/2017/07/collective-computation/533169/?utm_source=atltw overview here]) provides a useful lens for considering the ways increased machine learning tools might change market behavior. The first talk, given by Professor [https://www.santafe.edu/people/profile/jessica-flack Jessica Flack], will provide an overview of the collective intelligence phenomena. This 30 minute talk will be followed by a 30 minute group discussion. An important goal for the talk and the discussion is to illuminate why the market is itself a collective intelligence. The second talk, given by Professor [https://www.santafe.edu/people/profile/david-krakauer David Krakauer], will explore how machine learning tools have affected non-market collective intelligences. A key goal of this talk, and the subsequent discussion, is to look for patterns in the impact of machine learning tools on collective intelligences across three domains:<br/>
 +
# Interfaces and cognitive bottlenecks<br/>
 +
# Authority and homogeneity<br/>
 +
# Strategic, or game theoretic, behavior<br/><br/>
 +
Following the morning talks and discussions, there will be a 90-minute practitioner panel. The panel will begin with each participant offering their own insights (approx. 10 min each) as to how they believe machine learning tools are changing the behavior of markets. Subsequent moderated discussion with the panelists (approx. 25 min) and with the panelists and the whole room (approx. 25 min) will more deeply explore how the collective intelligence lens can help us understand the ways machine learning is changing market behavior.<br/><br/>
 +
'''Afternoon Sessions – Machine Learning in Markets and Increased Homogeneity'''<br/>
 +
We have already observed a host of mechanisms through which machine learning and other new technologies have affected financial markets (see [https://alo.mit.edu/wp-content/uploads/2017/06/Moores-Law-Vs.-Murphys-Law-in-the-Financial-System-Whos-Winning.pdf overview here]). The first talk, given by Professor [https://www.santafe.edu/people/profile/andrew-lo Andrew Lo], will provide an overview as to how these tools are changing market behavior. One notable mechanism involves increased homogeneity of market strategies. (As noted above, this is one of the canonical ways machine learning can affect collective intelligences.) In the second afternoon lecture, Professor [http://www.brandeis.edu/facultyguide/person.html?emplid=3863836c60fea8a993359c6d2f71be423bc77a23 Blake LeBaron] will use computational models to more fully explore the relationship between decreased strategic variety and market behavior. Following the afternoon talks and discussions, there will be a 90-minute practitioner panel. The panel will begin with each participant offering their own insights (approx. 10 min each) as to how they believe machine learning tools are affecting the behavior of markets. Subsequent moderated discussion with the panelists (approx. 25 min) and with the panelists and the whole room (approx. 25 min) will more deeply explore strategic homogeneity and other mechanism by which machine learning impacts market behavior.<br/>

Revision as of 20:36, 21 June 2019

Navigation


Asimov Robot.jpg

SFI ACtioN Topical Meeting


August 8, 2019
New York, New York


Morning Sessions - Collective Intelligence
The new field of collective intelligence (see The Atlantic’s overview here) provides a useful lens for considering the ways increased machine learning tools might change market behavior. The first talk, given by Professor Jessica Flack, will provide an overview of the collective intelligence phenomena. This 30 minute talk will be followed by a 30 minute group discussion. An important goal for the talk and the discussion is to illuminate why the market is itself a collective intelligence. The second talk, given by Professor David Krakauer, will explore how machine learning tools have affected non-market collective intelligences. A key goal of this talk, and the subsequent discussion, is to look for patterns in the impact of machine learning tools on collective intelligences across three domains:

  1. Interfaces and cognitive bottlenecks
  2. Authority and homogeneity
  3. Strategic, or game theoretic, behavior

Following the morning talks and discussions, there will be a 90-minute practitioner panel. The panel will begin with each participant offering their own insights (approx. 10 min each) as to how they believe machine learning tools are changing the behavior of markets. Subsequent moderated discussion with the panelists (approx. 25 min) and with the panelists and the whole room (approx. 25 min) will more deeply explore how the collective intelligence lens can help us understand the ways machine learning is changing market behavior.

Afternoon Sessions – Machine Learning in Markets and Increased Homogeneity
We have already observed a host of mechanisms through which machine learning and other new technologies have affected financial markets (see overview here). The first talk, given by Professor Andrew Lo, will provide an overview as to how these tools are changing market behavior. One notable mechanism involves increased homogeneity of market strategies. (As noted above, this is one of the canonical ways machine learning can affect collective intelligences.) In the second afternoon lecture, Professor Blake LeBaron will use computational models to more fully explore the relationship between decreased strategic variety and market behavior. Following the afternoon talks and discussions, there will be a 90-minute practitioner panel. The panel will begin with each participant offering their own insights (approx. 10 min each) as to how they believe machine learning tools are affecting the behavior of markets. Subsequent moderated discussion with the panelists (approx. 25 min) and with the panelists and the whole room (approx. 25 min) will more deeply explore strategic homogeneity and other mechanism by which machine learning impacts market behavior.