L’implication de Cambridge Analytica dans la campagne du Brexit était limitée… tout comme l’efficacité de ses outils

/l-implication-de-cambridge-analytica-da

  • L’implication de Cambridge Analytica dans la campagne du Brexit était limitée… tout comme l’efficacité de ses outils
    https://www.lemonde.fr/pixels/article/2020/10/08/l-implication-de-cambridge-analytica-dans-la-campagne-du-brexit-etait-limite

    Une enquête britannique conclut que l’entreprise au cœur du scandale de vol de données personnelles n’avait développé aucune technologie innovante. Cambridge Analytica « n’a effectué qu’un travail limité pour la campagne [pro-Brexit] Leave.eu, au-delà de son implication dans l’analyse des données des membres du parti UKIP ». C’est l’une des principales conclusions de l’enquête menée par l’Information Commissionner Office (ICO – « Bureau du commissaire à l’information ») britannique, qui a publié mercredi 7 (...)

    #CambridgeAnalytica/Emerdata #algorithme #manipulation #données #élections #profiling (...)

    ##CambridgeAnalytica/Emerdata ##publicité

    • @koantig c’est plus facile de penser que c’est du vent (et ça rassure la population de croire que ça n’a pas eu d’influence sur le vote vu la monstruosité du scandale) que d’évaluer à sa mesure le système de surveillance mis en place pour profiler les électeurs et les influencer dans leur vote. Plusieurs films documentaires sur Cambridge Analytica présentent des enquêtes extrêmement bien menées pour se faire une idée plus précise de cette surveillance et du basculement mené tout en maintenant le mythe démocratique du vote, cela permet de comprendre l’avènement de figure tel que Trump. Enfin, pour mieux comprendre la réification en bulletin de vote des êtres humains dans la sphère politique, on peut se faire une idée des forces en œuvre avec les côtes prédictives de la bourse mondiale et les gains pharaoniques et cyniques du même auteur du programme.

    • Je n’ignore évidemment pas les problèmes que la généralisation de la surveillance engendrent (j’écris ce message sur tor ! Et il n’y a qu’à voir le genre de trucs que je poste) mais il faut s’attacher à mesurer les effets de ces outils à leur juste valeur.

      Pour Trump et le Brexit, elles ont sans doute jouer un rôle mais c’est parce que le vote était déjà à 50-50. Un effet même marginal peut décider de l’issue. Le fait que ces élections étaient si serrées au départ est un problème plus profond.

      Le FUD et le confusionisme m’ont l’air beaucoup plus efficaces pour saloper la démocratie que des machines fonctionnant sous deep learning, nourri par la surveillance (raison de plus de s’en passer !).

    • Le mot Brexit n’apparait que 2 fois dans le rapport en ligne, j’ai repris les passages qui me semblaient intéressants. Je ne développerai pas parce que je trouve épuisantes et stériles les conversations « pour avoir raison ».

      https://ico.org.uk/media/action-weve-taken/2618383/20201002_ico-o-ed-l-rtl-0181_to-julian-knight-mp.pdf

      This further work confirms my earlier conclusion that there are systemic vulnerabilities in our democratic systems.

      The investigation is therefore concluding, and the following letter and Annexes acts as our final written account to Parliament. It provides a summary of the conclusions we have drawn from our analysis of the evidence in the final stages of our investigation, the additional actions we have taken and why, and broader learning we and other data protection authorities can draw on to inform future investigations and regulatory work in the digital era. In addition, Annex 1 provides the Committee with detailed answers to the specific questions asked by the Committee. Annex 2 provides a deep dive into how SCL Elections / Cambridge Analytica used the personal data it held, whether these methods could be used in the future, and the associated risks to citizens.

      The following organisations have now paid the penalty notices levied on them:•Facebook (£500,000) paid 04 November 2019•Vote Leave (£40,000) paid 29 April 2019•Leave.EU (£15,000) paid 15 May 2019 •Emma’s Diary (£140,000) paid 29 August 201814.In addition, we successfully prosecuted SCL Elections for their failure to comply with my Enforcement Notice. We fined them £18,000

      22 - What is clear is that the use of digital campaign techniques are a permanent fixture of our elections and the wider democratic process and will only continue to grow in the future. The COVID-19 pandemic is only likely to accelerate this process as political parties and campaigns seek to engage with voters in a safe and socially distanced way.

      23 - I have always been clear that these are positive developments. New technologies enable political parties and others to engage with a broad range of communities and hard to reach groups in a way that cannot be done through traditional campaigning methods alone. But for this to be successful, citizens need to have trust in how their data is being used to engage with them.

      27.The impact of this investigation has also had international reach. I have been asked to brief parliaments and governments across the world and I have shared the learning from this investigation with election oversight and privacy regulators internationally. The prominence of the use of personal data in political influence has grown significantly, and several international counterparts have since undertaken similar work, as is appropriate to safeguard their national democratic structures.

      –----

      10.SCL’s own marketing material claimed they had “Over 5,000 data points per individual on 230 million adult Americans.” However, based on what we found it appears that this may have been an exaggeration. 11.Although we do not have a list of all the datasets, during the document review we discovered evidence that some of the data sets as at September 2015 included:•Nationwide voter files from L2 (meaning “Labels and Lists”) and DataTrust (~50 data points for 160M individuals)•Nationwide consumer data from Acxiom and Infogroup (~500 data points for 160M individuals) •Election return results from Magellan (~20data points for national census tracks) •Nationwide consumer data from DataTrust (3000 data points for 100M individuals) •Psychographic inventories (10 data points for 30M individuals) •Facebook social network (graph database containing 30M individuals) •Facebook likes (570 data points for 30M individuals) •In-depth Republican Primary focused surveys (80k) •ForAmerica member data (14.6M post comments, 240M post likes across 31 M users) •Emails from Infogroup (30M) •Emails from DataTrust (26M)12.In short, the number of data points varied considerably, both from individual to individual and from one project to the next.

      14.In respect of Dr Kogan’s application, which he called thisisyourdigitallife(the App), the material obtained in the evidence review corroborated our understanding as set out in our previous reports that it obtained data from individuals who authorised it to access their Facebook data. However, the App functioned in a way which meant that it was also able to obtain the data of that user’s Facebook ’friends’ (who had not themselves restricted such sharing through their own Facebook ’privacy controls’). In conjunction with the personality quiz function of the App, along with a record of each user’s ’likes’ information, Dr Kogan was able to model personality traits for users of the App, and for their Facebook ’friends’. This approach seeming built on
      14earlier work by Dr Kogan involving Facebook ‘likes’ and personality scores.Dr Kogan set up a new company, GSR, this was established and funded for the primary purpose of acting as a vehicle for the provision of the services anticipated under the contract between GSR and SCL / CA.

      19.It was suggested that some of the data was utilised for political campaigningassociated with the Brexit Referendum. However, our view on review of the evidence is that the data from GSR could not have been used in the Brexit Referendum as the data shared with SCL/Cambridge Analytica by Dr Kogan related to US registered voters. There was evidence of considerable focus in the data collection and data matching processes between GSR and SCL on US voters, as this was what was to be paid for under the contract(s) between them. Cambridge Analytica did appear to do a limited amount of
      15workfor Leave.EU but this involved the analysis of UKIP membership data rather than data obtained from Facebook or GSR. Some evidence was recovered however that suggested an intention by SCL / GSR to target UK voters in 2014 through the same process. This work does not appear however to have been taken forward.

      23.In early 2014, SCL/CA commissioned Aggregate IQ ("AIQ"), a Canadian based company, to build a Customer Relationship Management (CRM) tool for use during the American 2014 midterm elections. SCL called the tool RIPON. It was designed to help political campaigns with typical campaign activity such as door to door, telephone and email canvassing. In October 2014, AIQ also placed online advertisements (including on the Facebook Platform) for SCL on behalf of its clients.

      29.The data points collected by GSR with respect to survey users and their Facebook ‘friends’ was specifically selected to enable a ‘matching’ process against pre-existing SCL databases. Matching took place using file sharing platforms and by reference to name, date of birth and location –with SCL’s existing datafiles being ‘enriched’ and supplemented by GSR’s data about those same individuals –and this matched information being passed back into SCL systems. This resulted for example information including scores for voting frequency, whether likely republican or democrat, voting consistency, and a profile which predicted personality traits matched to information such as voter ID, name, address, age, and other commercial data.

      30.Through such processes the relevant US voter GSR data (about approx. 30 million individuals) was then further analysed using machine learning algorithms to create additional “predicted” scores relating to partisanship and other criteria which were then applied to all the individuals in the database. Some of these focussed on likes as wide ranging as “gay rights”, “Obama the worst president in US history”, “Re-elect President Obama in 2012”, “the Bible” and “National Rifle Association”. These scores were used to identify clusters of similar individuals who could be potentially targeted with advertising relating to political campaigns. This targeted advertising was ultimately likely the final purpose of the data gathering but whether or which specific data from GSR was then used in any specific part of campaign has not been possible todetermine from the digital evidence reviewed. There is however evidence recovered that suggests that similar approaches and models based on the predicted personality traits and other measures were used with Republican National Committee (RNC) data.