Trajectories through Social Machines Kevin R. Page and David De Roure Oxford e-Research Centre, University of Oxford, UK.
Abstract A Social Machine is the consideration of a combined socio-
technical entity on the Web. Inherent in observing this composition is the need to apply mixed methods over both the technical infrastructure as we might see in a more traditional Web Observatory the social communities that interact with the Web, and their operation and interaction in combination as a Social Machine. To further complicate our observa-
tions, the intent of these constructs may dier both across concurrent social groups with divergent aims, and through evolution of an initial purpose over time. Such parallel and iterative use, re-use, and modi cation of social and computational elements confounds the application of a boundary de ning any single, static, Social Machine. To address the challenge of observing these ckle amorphous composites we introduce the notion of trajectories, as de ned by their purpose; trajectories provide a dimension through which we might study the interaction between the socio-technical elements. We identify distinct trajectories for concurrent purposes and for evolving purposes as they morph over time, and propose the comparative evaluation of such trajectories as a mechanism for better understanding the overall Social Machine.
1
Introduction
In observing Social Machines [1] we set out to analyse both computational (Web servers, clients, and other tooling) and human elements, considering them as a socio-technical whole of machines and people, rather than as strati ed layers of each. We have previously argued the utility of studying interactions between multiple Social Machines and the means by which they evolve and change over time individually and as a network; where elements may be reused within concurrent Social Machines beyond the intent of their original design and construction [2]. In that work we provided an illustration of The Machines of Spam which we revisit here for similar purpose. The scenario considers the evolving Social Machine(s) around the reCAPTCHA [3] service: 1. the genuine users of social websites who undertake reCAPTCHA mechanisms to prove they are human to those sites; 2. spam originators who procure the services of human mercenaries (via websites set up for this purpose), who in turn complete the reCAPTCHA to gain an account on the social websites, which is then traded with the spammer.
2
What to observe - whither the Social Machine?
In order to study and understand the emergence and behaviour of Social Machines, and indeed to construct the Observatory through which to undertake such study, an obvious rst step might be to de ne the limits of a Social Machine so it can be identi ed and observed. While this is an interesting intellectual question and certainly worthy of further consideration, the complexities and ever-changing nature of Social Machines (discussed in [2]) contrive to make this a diversionary exercise from that of observation.
Social Machine A
"Genuine" Social Website Users reCAPTCHA
Social Website
Social Machine C Social Machine B Mercenary Community Spam Originators
Website to coordinate mercenaries
Figure 1. Notional delimitations of Social Machines in the reCAPTCHA scenario (il-
lustrated as grey ellipses)
Figure 1 illustrates this conundrum for the reCAPTCHA scenario: since the reCAPTCHA service has been used by both genuine users and the the spammers' mercenaries it would seem of interest to de ne these Social Machines distinctly (Social Machines
A and B
respectively) and study the design and con-
struction of each as well as the interaction between them; yet at the same time we would like to consider the properties of an evolving emergent system exhibited
C ). There is, perhaps, no clear cut answer:
by the Social Machine as a whole (
1
all these aspects all these Social Machines
are worthy of investigation, but
without the clarity of borders the Social Machine(s) resist analysis as surely as they resist de nition.
1
In [2] a further Social Machine is identi ed, comprising the socio-technical elements of despamming scripts and websites listing known spammers. This too can can be represented as a trajectory, passing through the same and additional elements as the example in Section 3, but is omitted here for simplicity.
So while we can see there are many
potential
boundaries to our Social Ma-
chine(s), we should ask whether it is necessary to identify them before we build our Observatory? Are there alternative characterisations that might more clearly aid their study?
3
Trajectories
purpose as a key mechanism in the study trajectory 2 , each of which is de ned primarily by such a purpose; that is, an end for which a path
In our previous work[1] we proposed
of Social Machines. We extend that notion here to one of a through elements of the Social Machine is taken.
In Figure 2 we describe the same reCAPTCHA scenario, with the same Social Machine elements, but in terms of trajectories, each de ned according to a
i ii ) the subversion of the test by mercenaries on behalf of spammers;
purpose: ( ) the passing of the reCAPTCHA test to validate a user for genuine reasons; (
iii ) the use of human input to the reCAPTCHA tests for statistical training
and (
and solving of (optical) character recognition tasks a purpose more easily made explicit through this characterisation of trajectories. It is also worth noting that both the genuine and mercenary communities contribute jointly to the third trajectory, a subtlety similarly di cult to describe with an aggregating view of Social Machines.
Purpose: Pass test to join website "Genuine" Social Website Users
Trajectory i
reCAPTCHA
Purpose: Character Recognition
Trajectory iii Purpose: Pass test to vandalise website Mercenary Community Website to coordinate mercenaries
Spam Originators
Trajectory ii
Figure 2. Trajectories through the reCAPTCHA scenario and their distinct purposes
2
The word trajectory intended to impart the importance of the time axis in the study of Social Machines, as discussed in [2]
Having characterised and identi ed trajectories we can use them as an axis for study and comparison. A single trajectory collapses the dimensions of investigation to a single purpose at a single time; this provides a foundation for a clear and well de ned set of elements and assessment criteria along which qualitative and quantitative methods can be applied to inter- and intra-element behaviour. Furthermore, we can make comparison of social and technical elements (or sets of elements) as they transpose multiple trajectories, be that where trajectories comprise dierent elements due to diering purpose; or where the same elements present multiple trajectories over the passage of time (with potentially evolving purpose). For example, a comparative study of the reCAPTCHA service as it interacts in multiple trajectories might prove insightful in understanding the emergent properties of the wider Social Machine.
4
Summary and future work
In this paper we have proposed
trajectories as a mechanism through which the
key properties of Social Machines can be observed and studied: a dimension that captures the subtlety of interacting sub-Machines; enables distinction and comparison throughout an emergent Social Machine's lifecycle or lifespan; and respects consideration of both design and serendipity by way of purpose. Through trajectories we believe we can reduce the observed complexity of Social Machines, providing a path or sample along which mixed methods can be applied for comparative study, enhancing rather than compromising our ability to study the nature of Social Machines as a whole. We believe the identi cation of trajectories is critical to building Observatories for Social Machines and prioritising the properties that will provide most insight when measured within the Observatory. De ning measures than span a complete trajectory will clearly prove challenging; as a rst step we suggest comparative analysis of single (or small set of ) element
across trajectories may
help explore these boundaries.
Acknowledgements
We are grateful to all our colleagues for useful discussions and insights. This work is supported by SOCIAM: The Theory and Practice of Social Machines, funded by the UK Engineering and Physical Sciences Research Council (EPSRC) under grant number EP/J017728/1 and comprising the Universities of Southampton, Oxford and Edinburgh.
References 1. Tim Berners-Lee and Mark Fischetti. Weaving the Web: The original design and ultimate destiny of the World Wide Web by its inventor. Harper, 1999.
2. David De Roure, Clare Hooper, et al. Observing Sociam Machines Part 1: What to Observe? In Proc. 1st International Workshop on The Theory and Practice of Social Machines, 2013. In Press.
3. Luis Von Ahn, Benjamin Maurer, et al.
reCAPTCHA: Human-based character
recognition via web security measures. Science, 321(5895):14651468, 2008.
