28102020_OpportunitiesforAIinBorders_MS_conference_study
Dieses Dokument ist Teil der Anfrage „Study and presentation on AI“
Ref. Ares(2021)7038143 - 16/11/2021 European Commission: DG HOME Opportunities for Artificial Intelligence in the area of Borders, Migration and Security Borders Conference Presentation 28/10/2020
Context The aim of the study was to explore high-priority opportunities for using AI in process areas around Border Control, Migration and Security AI has become an area of strategic importance and a key driver of economic development, and as a consequence, the European Commission is developing an WHY? European approach to AI to ensure competitiveness and to shape the conditions for its development and use (ensuring respect of European values). The main objectives of the study are: • To identify a set of priority opportunities for using AI within Border Control, Migration and Security WHAT? • To transform these opportunities into a portfolio of initiatives, with roadmaps for specific implementation programmes The study focuses on 9 process areas related to Borders, Migration and Security. Border surveillance and WHERE? criminal investigations is out of scope, as also COVID-19 implications. © 2020 Deloitte Belgium 2
Scope of work The project look at nine processes, which were divided in external and internal processes IN SCOPE External processes Internal processes • Visa issuance process for short-term stay • EU policy making and enforcement process • Issuing ETIAS travel authorization • Operational management of systems and infrastructure • Issuance of docs for long-stay or residence in Schengen area • Inland control of migration status of TCN • Process for granting asylum or other measures • Consultation of SIS and involvement of SIRENE Bureaux • Border checks at the external Schengen borders NOT IN SCOPE • Areas covered by systems for border surveillance • AI for cyber-security • Police surveillance of public places and infrastructure • Data collected for law enforcement and criminal investigation • COVID-19 impacts and considerations © 2020 Deloitte Belgium 3
Methodology application The study followed a funnel-shaped approach with a broad identification phase followed by successive rounds of prioritisation, and is leading into initiative portfolio structuring and potentially PoC development Completed In progress IDENTIFY PRIORITISE REFINE FINALISE PoC 4 • Use cases built out into Proof-of- Identify Prioritise AI use Refine prioritised Build out detailed Concepts to opportunities to use cases based on use cases, to form a analysis and outline validate AI for meeting value and structured indicative roadmap expected value important needs within feasibility criteria opportunity portfolio for initial AI DG Home/Borders and feasibility domain • Subset of 4 use cases where selected as 1 2 3 candidates for • All use cases scored • Prioritised use cases combined into AI progressing • 9 processes • Filtered to ~70 initiatives with articulation of benefits, • One PoC is analysed • 35 use cases risks, etc. currently ongoing • 99 AI use prioritised in • Implementation roadmaps were at DG HOME cases identified workshop developed for each initiative © 2019 2020 Deloitte Belgium 4 © 2020 Deloitte Belgium
Outcome of Identification of AI Opportunities (“part 1”) Following the prioritisation workshop, we arrived at a balanced portfolio of use cases 5 Automation Portfolio Category 35 11 Engagement 16 7 12 Prioritised use cases Doing the same thing Doing the same thing Doing a different thing 19 Insight better differently Use cases, per use case group Use cases, per use case group Monitoring 1 Macro environment analysis 1 Internal chatbot 2 Integration fostering 1 Individual risk assessment 10 General risk assessment 1 Fraud detection 2 External chatbot 7 Ethics 1 Efficiency improvement 9 Chatbots, risk assessment and efficiency improvement amount almost 85% of the Most use cases prioritised fall under the ‘WOW’ quadrant. These are use cases with use cases identified. This can be explained by the wide application that this use high feasibility and high value. Nevertheless, some shorter and longer term use cases have throughout the different processes and its contextual importance. cases also fall under the ‘NOW’ and ‘HOW’ quadrants, respectivelly. © 2020 Deloitte Belgium 5
Use case example: Personalised visa application form & questions (VISA-9) Initiative 1: Visa issuance for short-stay Description Personalised application form, using AI to augment mandatory fields with a tailored list of questions for a specific individual's application, based on what will be most informative for the specific risk assessment at hand. Value drivers • Better analysis of applications by tailoring the questions to the individual, making them not unnecessary complex and to the point • Speed up and improve assessment by case workers, by providing an initial information gathering assessment from which a human can start from AI type Engagement © 2020 Deloitte Belgium Note: we will further dive into the details and discuss the applicability of several use cases after the break 6
Use case example: AI individual risk assessment (VISA-8) Initiative 1: Visa issuance for short-stay Description AI model to detect irregular travelling patterns (by consulting ETIAS/EES records) or applications. Could then either prompt a human to investigate further, or feed into an Engagement AI to ask the applicant for further information/documentation. Value drivers • Better insight in irregular travel patterns; • Faster identification of possible high risk individuals regardless of departing country; • Systematic check for possible irregular travel patterns which can prompt further investigation. AI type Insight © 2020 Deloitte Belgium Note: we will further dive into the details and discuss the applicability of several use cases after the break 7
Use case example: Computer vision to detect SIS alerts using cameras (SISSIRENE-1) Initiative 5: Consultation of SIS and involvement of the SIRENE Bureaux Description Use of computer vision detect SIS alerts using cameras - such as identifying a match on a target person or car number plate. Value drivers • Decrease the human effort in searching for people or objects • Improve the efficiency in the person or object identification • Increasing the security of the Schengen Area AI type Automation © 2020 Deloitte Belgium Note: we will further dive into the details and discuss the applicability of several use cases after the break 8
Outcomes of defining the programme of work (“part 2”) Programme of work clearly shows different themes in the upcoming years Stage PoC Pilot Roll out AI use case project Complexity Low Medium High 2 Application chatbots 7 8 (high feasibility) Computer vision(links from (6), advanced AI)) Advanced AI (more technical or 3 Risk assessment ethical challenges) (high value) 6 Other chatbots (follows (2) and links into (7)) 4 Knowledge mgmt. (technical link from (2) to (5) 1 Quick wins (high value & feasibility) 5 Policy NLP (high value, advanced NLP) CoE Initiatives Transversal Future innovation set up sponsored enablers set up enabled © 2020 Deloitte Belgium 9
Initiative 1: Visa issuance for short-stay Overview of the sequence and opportunities involved Chatbots Risk assessment Knowledge management Policy insight/analytics Computer vision VISA-1 VISA-3 VISA-8 VISA-9 Sequence • Technically • Requires a big amount of • Can be implemented • More “open-ended”: many straightforward: data: data quality is key as part of VISA-3 different levels of solutions for for the accurateness of complexity possible • Could both be Considerations development exist the model May require considerate implemented before • for sequencing • Relatively easy to • Data contains personal (as a PoC) or integration with systems develop along side info, which could be “plugged-in” after using personal data with the current visa blocking factor development of process VISA-3 • • • • Complexity ID Brief description Virtual assistant (or chatbot) supporting visa application process by (1) taking in information, (2) answer questions posed by the applicant and (3) ensure VISA-1 data quality VISA-3 Application triaging using individual risk assessment for rapid, more efficient risk analysis VISA-8 Identification of irregular travelling patterns VISA-9 Use of a personalised application form using AI to tailor questions asked to the applicant creating an augmented application form. 10