Ceci est une version HTML d'une pièce jointe de la demande d'accès à l'information 'EPSO-Databases'.


Ref. Ares(2015)4498011 - 22/10/2015
From: 
 (SG) 
Sent: 
10 February 2013 13:42 
To: 
 (EPSO); 
 (EPSO); EPSO ACCES 
DOCUMENTS 
Cc: 
 (EPSO); 
 (SG); SG DOSSIERS ACCES 
Subject: 
FW: New confirmatory request for access to documents 
Attachments: 
FINAL 29.01.2012 .pdf; EU
 draft 8.2.2013.doc 
Categories: 
 
Dear 
 
I hope you are well. I have been myself absent during January for health reasons. 
I have examined the reply given by EPSO to the new request of Mr 
.He introduced his 
application because of the lack of initial reply . He has contacted us today to confirm his interest in his 
application after receiving EPSO position.  
Consequently, please find enclosed a draft reply including SG questions as well as a copy of the definitive 
reply given by SG to his previous confirmatory request. The reply to his previous request has received 
EPSO agreement but it could not been sent before since we have only received the approval of the LS last 
week.  Although EPSO reply confirms our position regarding his previous request, there are some issues 
that I would like to discuss with you before launching the LS consultation. 
 
 
Surprisingly, he stated in his message (see copy enclosed) that he understood that the document 
requested did not exist, which is precisely the Commission conclusion that he did not want to accept. He 
also recognizes that his confirmatory application modified his initial request. Both are the main points we 
address in our reply to his previous request.  
SG will like to close this new request as soon as possible. Therefore,I  would appreciate to have your views 
by Tuesday 12.02 2013 (end of business) 
Thanks in advance 
 
From: EPSO ACCES DOCUMENTS  
Sent: Tuesday, February 05, 2013 5:18 PM 
To: SG DOSSIERS ACCES 
Cc: 
 (EPSO) 
Subject: RE: New confirmatory request for access to documents 
Cher M. 
 
















Afin de répondre à votre demande, je vous informe que : 
 
1) En ce qui concerne les documents refusés, la gestionnaire du dossier étant en congé de 
maladie, voudriez-vous la contacter lundi 11/02. 
2) Mme 
 (
) est gestionnaire du dossier au stade de la demande initiale. 
3) La réponse à la demande initiale est attachée dans Gestdem. 
 
 
 
Unité 06 : Communication et attractivité
www.eu-careers.eu 
QUESTIONS JURIDIQUES 
Facebook: EU Careers
Secrétariat du secteur juridique 
Twitter: EU_Careers 
Tel. 
 
European Personnel Selection Office
 
 
 
 
 
 
From: SG DOSSIERS ACCES  
Sent: Wednesday, January 30, 2013 6:36 PM 
To: EPSO ACCES DOCUMENTS 
Subject: FW: New confirmatory request for access to documents 
 
Cher(e) Collègue,  
Je vous prie de bien vouloir trouver ci-joint, copie de la lettre que M. 
 
a envoyé à la Secrétaire Générale, par laquelle un recours est introduit pour 
absence de réponse 
à certains documents demandés à la DG EPSO (GESTDEM n° 
2013/68
). 
Je me permets de vous rappeler que le délai impératif de réponse est de 15 jours 
ouvrables à partir de l'enregistrement de la demande. 
Afin de permettre à la Secrétaire générale de répondre à ce recours avant 
l'échéance  du 20/02/2013  prochain,  je  vous  remercie  d'avance  de  bien  vouloir  me 
communiquer (via l'adresse électronique SG DOSSIERS ACCES) immédiatement (dans les 
24 heures – délai:01/02/2013) : 
1)      les documents refusés lors de la demande initiale, afin de nous permettre une 
évaluation indépendante de la pertinence du refus (si les documents sont trop 
nombreux, nous vous demandons de nous faire parvenir au moins une liste des 
documents concernés);  
 
2)   le nom du (de la) gestionnaire du dossier au stade initial.
 



Dans les jours suivants, un(e) gestionnaire (administrateur) de l'équipe "accès aux 
documents" du SG prendra contact avec le (la) gestionnaire du dossier au niveau de la 
DG pour préciser les éléments complémentaires de contribution dont le SG aura 
besoin. Une prompte réponse à cette démarche sera attendue. 
Pour toute information complémentaire, vous pouvez prendre contact avec moi. 
Merci d'avance de votre collaboration. 
 
 
 
SG.B.5.  
Transparence. 
Berl.

 
 
 
From: SG ACCES DOCUMENTS  
Sent: Wednesday, January 30, 2013 6:33 PM 
To: 
 
Subject: RE: New confirmatory request for access to documents 
 
Dear Sir,  
Thank you for your e-mail dated 30/01/2013, registered on 30/01/2013 
[Ares(2013) 119295]. 
I hereby acknowledge receipt of your confirmatory request for access to 
documents - Gestdem 2013/68. 
In accordance with Regulation 1049/2001 regarding public access to 
European Parliament, Council and Commission documents, you will receive a 
response to your request within 15 working days. 
 
Yours sincerely, 
 
 
SG.B.5.  
Transparence. 
Berl.

 
 

 
From: 
 [mailto:
]  
Sent: Wednesday, January 30, 2013 8:21 AM 
To: SG ACCES DOCUMENTS; 
 (EPSO); 
 (SG) 
Subject: New confirmatory request for access to documents 
 
Dear Secretary General, dear Mr. 

 
 on December 28, 2012, I submitted a request for access to 
documents under Regulation (EC) No. 1049/2001 to EPSO. 
On January 8, EPSO acknowledged receipt - EPSO-06/RK D(2013) 
20023 - and stated that the time limit for handling the application 
expires on January 29, 2013. 
EPSO failed to reply within the prescribed time limit. 
This entitles me to making a confirmatory application according to 
Article 7, Paragraph 4, of Regulation (EC) No. 1049/2001. Please 
find this confirmatory request, which is identical to my original 
request from December 28, 2012, below. 
 
Please acknowledge receipt of this confirmatory request
 
This is a new confirmatory request for access to documents under 
Regulation (EC) 1049/2001 which differs from my previous 
confirmatory request (from August 21, 2012) both in scope 
and because I now do not ask for a single table but for parts of 
multiple existing documents. 
 I realize this request is a very long and complicated document. The 
second-to-last paragraph contains 6 simple steps for fulfilling this 
request, with two of them probably not even needed (No. 1 and No. 
6). So, in case you just want a quick overview of the request and of 
how you could easily grant it, I recommend reading the second-to-
last paragraph with its 6 steps first. I had to add all the other 
paragraphs to make the request fit Regulation (EC) No 1049/2001. 
 REQUEST FOR ACCESS TO DOCUMENTS 
I request anonymized answer data from the admission test for the 
open competition EPSO/AD/230/12 (AD 5) AND 
EPSO/AD/231/12 (AD 7). 
Please send me, in electronic form, a set 
of documents (existing documents or parts thereof)
 from which I 
will be able to construct a table with one row for every 
candidate/question pair in the admission test (if there were 45357 

candidates and 60 questions, this table should contain 2,721,420 
rows), and with the following columns: 
•         Candidate ID: an identifier of the candidate. This identifier 
should not give me any indication of the identity of the 
candidate who gave this answer, but it should be the same 
for all answers by the same candidate, and no two 
candidates should share the same candidate ID. For 
anonymization purposes, it is completely acceptable for me 
if you hash the Candidate ID, for instance by using the 
MD5 function, as long as you do it consistently in all 
released documents in which it appears so that the links 
across documents are not destroyed. 
•         Question ID: an identifier of the question, not the question 
itself. This identifier should allow you to identify the 
question, but it should not give me any indication of the 
question's contents. The same question in two different 
languages should be listed under the same question ID in 
the table I need to be able to construct. In the documents 
you give me, the information which questions are 
translations of the same original question can of course be 
present in whatever form in which you currently store it, not 
necessarily in the form of a common question ID for all 
translations of the same original question. 
•         Question Type: an indicator of whether this question is a 
verbal reasoning, abstract reasoning, numerical reasoning, 
or situational judgement question. 
•         Language: the language in which this question was 
presented to this candidate. 
•         Neutralized: the information whether the answers to this 
question were "neutralized" or not. By "neutralized", I mean 
that any answers to this question were disregarded and 
instead scores from the non-"neutralized" questions were 
extrapolated. I do not imply here that a retroactive 
neutralization which deserves this name is actually possible. 
•         Expected answer: an identifier of the answer that is 
expected. This identifier should not give me any indication 
of the contents of the answer, but for the same 
question/answer pair, it should stay the same. If the answer 
options were always presented in the same order, the letter 
A/B/C/D would be sufficient. If answer options were not 
presented in the same order to all candidates, the same 

identifier should always be used for the same answer 
option. For situational judgement questions, the entire 
expected answer should be indicated, that is, the best and 
the worst option. Again, since I cannot know how you store 
this information, I request it in whatever form you have it. 
•         Given answer: an identifier of the answer chosen by this 
candidate for this question. Again, I do not request any 
indication of the contents of the answer, just an identifier. It 
should be specified in the same fashion as for the expected 
answer, so that correct answers can be recognized by the 
identity of identifiers. If the candidate did not answer a 
question, this information should be included as well. For 
situational judgement questions, the complete answer (that 
is, best and worst option). 
•         Time spent: the time (for instance, in seconds) the 
candidate spent on this question. 
•         Difficulty: the level of difficulty of this question. 
I do not request that you create a new document by merging 
information from existing documents, but instead I request that you 
provide me with the existing documents which are stored in 
electronic form in your various IT systems, possibly after removing 
some information from them which I did not request and which is 
covered by one or more exceptions from Article 4. 
It is, as far as I know, not publicly known how many documents 
you are using to store the information contained in the table which I 
want to be construct from your documents, so I cannot specifically 
ask for individual documents about whose existence I can be 
certain. However, you have already confirmed that the information 
I requested on July 2, 2012 exists (EPSO/RK D(2012) 1130795: 
"EPSO does indeed possess the information referred to in your 
request"). That is, the existence of all requested information except 
for the difficulty column has already been confirmed. 
  
Although I do not know and cannot know how you store this 
information about which we agree that it exists, let me make an 
educated guess about which documents exist. If my guess is 
correct, please deliver those documents. If it is not correct, please 
deliver a set of documents containing equivalent information. 
  

You probably have a document (I will call it "candidate table") 
containing a list of all candidates and several properties of the 
candidates. In other documents, you probably have 
pointers to records in this candidate table. These pointers use one 
column (or a column combination) in the candidate table which I 
will call the "Candidate ID". In database terminology, this 
Candidate ID would be called a key. 
I request the following subset of the candidate table: 
All rows with candidates who took part in the competition 
EPSO/AD/230/12 (AD 5) and/or EPSO/AD/231/12 (AD 7). Each 
row should contain at least these columns: 
•         Candidate ID (the ID used in other 
documents to refer to records of this document; if multiple 
such IDs exist, I request all of them; if the IDs would reveal 
the identity of candidates, I request that you anonymize this 
information without losing the identifying property of the 
ID so that the references from the other documents still 
work; such anonymization would constitute the removal of 
information from a document to arrive at a part of a 
document in the sense of Regulation (EC) No 1049/2001. 
Art. 4.6: "If only parts of the requested document are 
covered by any of the exceptions, the remaining parts of the 
document shall be released." The identity of a candidate is 
probably covered by one of the exceptions, but the 
information which answers have been given by the same 
candidate is not. So you should not release the identity of 
candidates to me, but an anonymous identifier is still 
needed for releasing the information which answers come 
from the same candidate). 
•         language combination of this candidate, in other words, 
information which will allow me to determine, in 
conjunction with the question type, which language was 
used for which question when it was presented tothis 
candidate. 
Please remove any additional columns which would trigger an 
exception according to Article 4. 
  
You most likely have a document (I will call it "question table") 
containing a list of all questions, answer options, expected answers, 
and difficulty. In other documents, you probably have 

pointers to records in the question table. These pointers use one 
column or column combination in the question table which I will 
call "Question ID". 
I request the following subset of the question table: 
All rows pertaining to questions which were used at least once in 
the competition EPSO/AD/230/12 (AD 5) and/or EPSO/AD/231/12 
(AD 7), with at least the following columns: 
•         Question ID (the ID used in other 
documents to refer to records of this document; if multiple 
such IDs exist, I request all of them; if the IDs would reveal 
the contents of the question, I request that you anonymize 
this information without losing the identifying property of 
the ID so that the references from the other documents still 
work). The same question in two different languages should 
be listed under the same question ID, or if this is not how 
you store this information, you should give me this 
information (which questions have the same content in 
different languages) in whatever format you use. 
•         Difficulty: the level of difficulty of this question 
•         Question Type: an indicator of whether this question is a 
verbal reasoning, abstract reasoning, numerical reasoning, 
or situational judgement question. 
•         Neutralized: the information whether the answers to this 
question were "neutralized" or not. See above for a 
definition of what I mean by "neutralized". 
•         Expected answer: an identifier of the answer that is 
expected. This identifier should not give me any indication 
of the contents of the answer, but for the same 
question/answer pair, it should stay the same. If the answer 
options were always presented in the same order, the letter 
A/B/C/D would be sufficient. If answer options were not 
presented in the same order to all candidates, please ensure 
that the same identifier is always shown for the same 
answer option. For situational judgement questions, indicate 
the entire expected answer, that is, the best and the worst 
option. 
Please remove any additional columns which would trigger an 
exception according to Article 4. 

  
You most likely have a document (I will call it "answer table") 
containing a list of all answers given by candidates in the 
competition EPSO/AD/230/12 (AD 5) and/or EPSO/AD/231/12 
(AD 7). 
I request the following subset of the answer table: 
All rows which were entered for the competition EPSO/AD/230/12 
(AD 5) and/or EPSO/AD/231/12 (AD 7), with at least the following 
columns: 
•         Candidate ID: an identifier of the candidate. This identifier 
should not give me any indication of the identity of the 
candidate who gave this answer, but it should be the same 
for all answers by the same candidate, and no two 
candidates should share the same candidate ID. The 
candidate ID should allow me to join this table with the 
candidate table. See also my other remarks about Candidate 
IDs above. 
•         Question ID: an identifier of the question, not the question 
itself. This identifier should allow you to identify the 
question, but it should not give me any indication of the 
question's contents. The same question in two different 
languages should be listed under the same question ID (also 
see my remarks about the question ID above). The question 
ID should allow me to join this table with the question 
table.  
•         Given answer: an identifier of the answer chosen by this 
candidate for this question. Again, I do not request any 
indication of the contents of the answer, just an identifier. It 
should be specified in the same fashion as for the expected 
answer, so that correct answers can be recognized by the 
identity of identifiers. Also, this identifier should not only 
contain the information whether the question was correctly 
answered or not, but also which wrong answer was selected 
in case the answer is wrong. In other words, for any pair of 
candidates who chose the same wrong answer option, this 
fact should be visible from this identifier, even though the 
answer itself should not be released to me. A separate 
identifier should be used if the candidate did not answer this 
question. For situational judgement questions, indicate the 
complete answer (best and worst option). 

•         Time spent: the time (for instance, in seconds) the 
candidate spent on this question. 
Please remove any additional columns which would trigger an 
exception according to Article 4. 
  
If you have split any of the requested documents in multiple 
documents (either horizontally, that is, by maintaining separate sets 
of rows, or vertically, that is, by maintaining separate sets of 
columns, or both), I request all documents which I 
need to construct the table which I have described first in my 
request. 
In the unlikely case that in one or more of the requested documents, 
there are no ID columns which identify records without giving 
away important content such as the name of a candidate, his 
passport number, or the wording of a question or answer, and you 
try touse this as an argument for not disclosing requested data, I 
request that you apply a hash function such as MD5 to the affected 
IDs in order to remove the information which should not be shared 
while retaining the information that must be shared (the "remaining 
parts" of the document include the information which records from 
one document need to be joined with what records from another 
document). 
  
Please note that I do not request any processing of data beyond 
what is required by Regulation (EC) 1049/2001. I only request the 
releasable parts of existing documents. Any data processing I ask 
for is only needed for separating the releasable information from 
information which would trigger an exception according to Article 
4, and Article 4 obliges you to perform this minimal amount of data 
processing. Art. 4.6 says that "If only parts of the requested 
document are covered by any of the exceptions, the remaining parts 
of the document shall be released." In other words, you are 
obliged to separate the remaining parts of the documents from the 
non-releasable parts of the documents if the documents contain 
non-releasable parts. If you store any requested information in a 
format different from what I describe here, I do not request re-
formatting – giving it to me in the existing format would satisfy my 
request, as long as equivalent information can be extracted from 
what you deliver. 
 

QUICK OVERVIEW OF HOW TO FULFIL THE REQUEST 
WITH MINIMAL EFFORT 
 
Please note that you can fulfil my request quite easily 
by slightly changing your existing mechanism for reporting the 
candidates’ answers back to them. The table I want to ultimately 
construct (see above) differs only slightly from the concatenation 
of tables which you already routinely send to every candidate as 
part of the application process, similar to the message you sent me 
via my EPSO account on 28/06/2012 (Candidate number: 
2575906). I realize that following these 6 steps could be construed 
as creating a new document which does not yet exist. Because of 
this, I do not request that you follow these 6 steps in particular. I 
merely point out that this would probably be the most efficient and 
economical way of fulfilling my request for parts of existing 
multiple documents outlined above, and I would also accept such a 
new, not yet existing document instead of the requested parts of 
existing documents in case you would prefer delivering the 
requested information in this equivalent form. 
 
You could deliver all requested information by modifying your 
existing data export mechanism with the following simple steps: 
1. Optional: anonymize the existing Candidate ID in case you 
consider it too revealing. This can be trivially achieved by 
supplying the ID in hashed form, for instance, by supplying 
md5(candidate ID) instead of the raw candidate ID. 
See http://en.wikipedia.org/wiki/MD5 and http://php.net/manual/en
/function.md5.php 
2. Combine the language combination and the table with 
expected/given answers. Since you already send them both within 
the same message (e.g., in your message to me from 28/6/2012), 
this should not be much effort. 
3. Add the difficulty. This information must be stored together with 
the question in the database of questions (or at least linked to it) 
and should therefore be easily retrievable. 
4. The question ID would have to be given to me as a globally 
unique ID (i.e., as an identifier that stays the same across all 
candidates), not just a candidate-specific question number. Since 
your mechanism for generating the tables you send to candidates 
clearly has a link between the global question ID and the candidate-
specific number (which must be true because you report the 
expected answer), reporting this global question ID in 
addition to the candidate-specific question number is a trivial 
change.   

5. Add the situational judgement questions to the mechanism 
already in place for all the other question types. 
6. Only if the answer options are presented to different candidates 
in different orders, you might have to compensate for this. 
  
WHY AM I ENTITLED TO RECEIVING THIS FILE FROM 
YOU? 
•         Regulation 1049/2011 applies because the file specified 
above fits the definition of a document in Art. 3a, I am a 
citizen of the Union residing in a Member State, and the 
document is held by an institution. 
•         No exceptions according to Art. 4 apply: 
Art. 4.1(a) does not apply because public security, defence 
and military matters, international relations, and 
financial/monetary/economic policy are not affected. 
Art. 4.1 (b) does not apply because the data I request are 
anonymized. 
Art. 4.2 does not apply because the contents of questions 
and answers are not included in the data I request. 
Art. 4.3 does not justify an exception because disclosure 
would not undermine the ongoing decisionmaking process, 
and in any case there is an overriding public interest in 
disclosure (transparency, ensuring fairness, the public 
needs to be able to check whether, and to what degree, 
language-based discrimination occurs). 
The rest of Art. 4 obviously does not justify an exception 
either. 
 
With best regards, 
 
 
 
 
-- 
 
 
 
Tel 
 
Fax 
 

Document Outline