Actually, the Kindle polls Amazon servers for a so-called "todo" list, which usually includes items like books or samples to download or bookmarks to update. However, it can also include a "send log" command which makes Kindle upload the system log to Amazon. Among other things, that log includes filenames and titles of the books opened by the user.
An example from my old investigation:
Code:
<?xml version="1.0" encoding="UTF-8" ?>
<response>
<total_count>5</total_count>
<next_pull_time>20080124 16:17:30</next_pull_time>
<items>
<item priority="0" type="CMND" action="SND" key="1199422466841:LOGFILE_MESSAGES_LONG" is_incremental="false" sequence="0">CustomerService</item>
<item priority="0" type="CMND" action="SND" key="1199422466841:611" is_incremental="false" sequence="0">CustomerService</item>
<item priority="0" type="EBSP" action="GET" key="B000WDV7TS" is_incremental="false" sequence="0">Chemoton Theory: Theory of Living Systems (Mathematical and Computational Chemistry)</item>
<item priority="5" type="NAMS" action="GET" key="Register" is_incremental="false" sequence="0" />
<item priority="5" type="PSNL" action="GET" key="ThankYouLetter" is_incremental="false" sequence="0">Thank You Letter</item>
</items>
</response>
(note the CMND (command) SND (send) items)
A piece of the system log:
Code:
Nov 20 12:26:51 syslog: info: Switched to Booklet: Bookworm
Nov 20 12:26:51 syslog: info: Book Title The Washington Post
Nov 20 12:26:51 syslog: info: Book AUTHORS The Washington Post Company
Nov 20 12:26:51 syslog: info: Book ASIN B000ZFV9QO
<....>
Nov 22 16:32:08 syslog: info: Search: Indexing content of /mnt/us/documents/Grantville_Gazette_Volume_I.prc at 16:32:08
Nov 22 16:32:09 syslog: info: Search: Sorting book at 16:32:09
Nov 22 16:32:35 syslog: info: Switched to Booklet: Browser
Nov 22 16:33:19 syslog: info: Switched to Booklet: Browser
Nov 22 16:34:32 syslog: info: Search: Writing index at 16:34:32
Nov 22 16:36:33 syslog: info: Search: 265.01 seconds to index the content of /mnt/us/documents/Grantville_Gazette_Volume_I.prc
Nov 22 16:36:35 syslog: info: Search: Content index needs to be generated because titleInfo is not valid: /mnt/us/documents/The_Tyrant.prc
I don't know whether Amazon actually uses that information or not (my guess would be they care only about diagnostic info), but they certainly can get it.