Research Internship

Between January and March 2014, I undertook a residential research internship at Blast Theory, a world renowned interactive performance and arts group based in Brighton, UK.

Rider Spoke Project

The purpose of this internship was to assist the Blast Theory team of artists in the development of a future project that engages with user profiling and personalisation.

Many of our everyday activities, experiences and interactions with technology engage with some level of user profiling: capturing and understanding the personal and behavioural qualities of these users and their individual trajectories through life. This information can then be used to personalise the engagements they have with various services and experiences. Alongside the commercial industry, artists are now increasingly looking towards profiling and personalisation to deliver artistic or performance interactions, which the aim of enhancing the user’s experience. Blast Theory have a long tradition of experimentation and engagement with socio-political performance art and the current global concerns over privacy and ownership of personal data appear like a natural area for these artists to engage.

Blast Theory’s past work historically participates with profiling (ex. Prof Tanda & ivy4evr) and their intention appears to be a continuation along this path. Therefore, they were keen to understand the current ‘state’ of the research within this area and also recognised that many other artists, academics, broadcasters and curators are equally engaged with this topic. In response to this, Blast Theory had partitioned their annual conference Act Otherwise to explore, discuss and reflect upon issues of personalisation and profiling, with a range of professionals from the areas listed above. The Act Otherwise – The invisible Hand: On Profiling and Personalisation conference took place on 27th & 28th February 2014, at the end of my internship.

Act-Otherwise-2014-14_chosen-1920x1080

Blast Theory artist Nick Tandavanitj supervised my residency and was keen to unearth some information, practices and outcomes of technical solutions for data mining and user profiling. Nick was particularly interested in the distinction between types of data mining, for example:

Participatory sensing:  Requires a user’s active participation in the collection data.

Opportunistic sensing:  Automated sensing, no user intervention. One of the main challenges of using opportunistic sensing is the phone context problem; for example, the application wants to only take a sound sample for a city-wide noise map when the phone is out of the pocket or bag. These types of context issues can be solved by using the phone sensors; for example, the accelerometer or light sensors can determine if the phone is out of the pocket.

Personal sensing:  Applications designed for a single individual that are often focused on data collection and analysis.

Community sensing:  Applications or systems for groups of users.

We explored a range of academic research and commercial mobile applications, which use mobile sensors such as accelerometer, microphone, GPS and ambient light sensors to capture user behaviours. For example, some research and commercial smartphone applications have been developed for the purpose of predicting and identifying transport modalities such as being stationary, walking, running, ascending and descending stairs, cycling, as well as different modes of motorised transport (cars, bus and trams). A range of technical processes have been explored, which used one or a combination of accelerometer, magnetometer, GPS, WiFi and GSM Cell Tower data streams to make predictions. Reddy et al. [1] devised a process that combined both accelerometer and GPS data streams to recognised fine grained modes of transportation. They observed similar accelerometer profiles when users were still, biking or in engaged in motorized travel. Whereas biking, running and walking all displayed similar GPS profiles. In situations where accelerometer profiles were similar, the distinguishing feature between them was speed, which could be more accurately indicated by GPS data. Therefore, a combination of the two data profiles successfully differentiated between a wider range of transport modalities than was found when using either the accelerometer or GPS sensor in isolation [1]. Hemminki et al. [2] showed that the location of a smartphone and its orientation did not deter the accelerometer from capturing the profile of a mode of transport (in this instance, on trams). Kwapisz et al. [3] demonstrated it was relatively easy to determine whether someone was sitting or standing using as the primary differences between these activities is the relative magnitudes of values for each axis, due to the different orientations of the device with respect to the Earth. They were able to determine sitting accuracy of 95.7% and standing 93.3%. These examples outline the potential of a user’s mobile phone powerfully communicating when someone is in transit and what type of transportation they may be using. When combined with other information such as a user’s point of origin it could be possible to recognise which bus, train, or road they are travelling on and where that user’s destination may lie. Some existing commercial smart phone applications already integrate some of this functionality such as Moves and Exist. They monitor user movement and behaviours to log activity, primarily for health promotion.

Other research explored during this time concerned data mining of social network sites (SNS) and in particular Facebook. Two key research projects were discovered, both of which aimed to identify a user’s personality type and behavioural characteristics by analysing either their Facebook Likes, or use of vocabulary on Facebook status updates. myPersonality is a long running project from Cambridge University, UK. Their primary study suggested that a rich picture of personality can be drawn from Facebook ‘Likes’. They found that they could automatically and accurately predict a range of personal attributes including: sexual orientation, ethnicity, religious and political views, personality traits, intelligence, happiness, use of addictive substances, parental separation, age, and gender. The research was less successful in determining personality types. Data from 58,466 US Facebook users who chose to engage in the project was analysed. In addition to giving permissions to their Facebook profile participants also completed the following surveys: the International Personality Item Pool (IPIP) questionnaire, which assess the ‘big five’ personality traits (openness, Conscientiousness, Extraversion, Agreeableness, Emotional stability), Raven’s Standard Progressive Matrices test to measure intelligence, Satisfaction with Life test [4]. As part of their ongoing research, the website http://youarewhatyoulike.com offers personality type analysis from your own Facebook ‘Likes’, the results of which are compared to your Facebook friends. Schwartz et al. [6], using the same data set, chose to analyse 700 million words, phrases, and topic instances collected from the Facebook status updates of 75,000 volunteers. They performed an open language analysis to identify the dominant words and phrases that most distinguished each of the Big Five personality traits. International Personality Item Pool (IPIP) questionnaire Linguistic Inquiry and Word Count (LIWC) software used to analyse text Correlated personality traits with online word usage. They found striking variations in language across personality, gender, and age. For example, subjects living in high elevations talk about the mountains; neurotic people disproportionately use the phrase ‘sick of’ and the word ‘depressed’; active life implies emotional stability; males use the possessive ‘my’ when mentioning their ‘wife’ or ‘girlfriend’ more often than females use ‘my’ with ‘husband’ or ’boyfriend’.

We also stumbled upon a Python script written by Keith Lee, an analyst for Trustwave, a company engaged in assessing security risks [7]. The script, called FBStalker (Facebook Stalker) is an open source intelligence (OSINT) gathering tool that reverse-engineers the Facebook Graph to reveal information contained on that users profile. If the user profile is secured, it searches for images, status updates or other media tagged with that user’s name on a friends open profile and then mines profile information from that friends account. A tandem script entitled GeoStalker uses location data to search for posts or media tagged to that location. For example, after inputting a set of coordinates, GeoStalker  then searches sites such as Flickr, Foursquare and Facebook for any media or post tagged with that location. The results then provide a method of accessing profile information about any user who uploaded, or happened to be tagged to those coordinates.

These research themes, conducted through the course of this internship, revealed a wide range of interesting research and commercial activity within data mining and user profiling. They sought to explore a range of technical methods and data analysis to expose meaningful approaches to data mining.

During the Act Otherwise – The invisible Hand: On Profiling and Personalisation Conference, I supported Blast Theory with preparations for the event and managed the live streaming of the event, which was hosted on the Blast Theory website. Additionally, I had prepared a presentation planned for the conference, which covered elements of the research I had undertaken, but unfortunately time restraints prevented me from delivering the presentation.

My time with Blast Theory was particularly engaging, especially witnessing the passion and precision with which these artists approach their work. The Act Otherwise Conference provided an equally stimulating and vibrant exploration of profiling and personalisation from many different perspectives. I am looking forward to following up some of the contacts I made at this event, who are engaged in areas of work related to my own. Finally, I look forward to engaging with future Blast Theory performance works.

 


References:
[1]          Reddy, S., Mun, M., Burke, J., Estrin, D., Hansen, M., & Srivastava, M. (2010). Using mobile phones to determine transportation modes. ACM Transactions on Sensor Networks (TOSN)6(2), 13.
[2]          Hemminki, S., Nurmi, P. and Tarkoma, S. 2013. Accelerometer-based transportation mode detection on smartphones. In Proceedings of the 11th ACM Conference on Embedded Networked Sensor Systems (SenSys ’13). ACM, New York, NY, USA, , Article 13 , 14 pages. DOI=10.1145/2517351.2517367 http://doi.acm.org/10.1145/2517351.2517367
[3]          Kwapisz, Jennifer R., Gary M. Weiss, and Samuel A. Moore. “Activity recognition using cell phone accelerometers.” ACM SIGKDD Explorations Newsletter 12, no. 2 (2011): 74-82.
[4]          http://www.mypersonality.org [Accessed Jan 25th 2014]
[5]          http://www.youarewhatyoulike.com [Accessed Jan 25th 2014]
[6]          Schwartz HA, Eichstaedt JC, Kern ML, Dziurzynski L. (2013) Personality, Gender, and Age in the Language of Social Media: The Open-Vocabulary Approach. PLoS ONE 8
(9): e73791.doi:10.1371/journal.pone.0073791 http://www.plosone.org/article/info:doi/
10.1371/journal.pone.0073791
[7]          https://www.github.com/milo2012/osintstalker [Accessed February 1st 2014]

Advertisements

One Response to Research Internship

  1. Pingback: Blast Theory Internship Residency | Adrian Hazzard

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s