Methodology

02Artificial intelligence

Bill summaries ("TL;DR") are generated automatically by a language model from the official text of the bill - they are always marked as such on the bill's page, and the full text is the binding one.

A language model is also used at two stages of the positions pages and the quiz: to identify which side of a disputed question a published text takes (a candidate's platform, statement or tweet), and to determine which side of the question a "for" vote on a bill supports, based on the bill's official summary. The model does not see who said the words or how anyone voted; it must quote the passage it relies on (and the quote is checked automatically against the source, otherwise the finding is discarded), and "unclear" is a legitimate answer. Every decision is stored and open to review. In addition, a model tags tweets in order to select the rounds in the "Whose tweet is this?" game.

No numerical figure (votes, seats, attendance and so on) is calculated or invented by artificial intelligence - every number is derived directly from its source.

03Voting attendance

The percentage of main votes in which the Member of Knesset took part (for, against or abstain) out of the votes in which they were eligible to vote - that is, only votes held during their actual period of membership in the Knesset, not the whole term.

A minister without a seat (under the Norwegian Law) is not included in the attendance calculation - they have no votes to miss. Attendance is shown as "Not enough data" for anyone who entered recently and does not yet have enough relevant votes for a meaningful calculation.

04"Bills initiated"

The count is based on who is listed as the bill's primary initiator (not co-initiators), according to Knesset data. A bill with one primary initiator counts once for that person; a bill with several "primary" initiators (a formal joint initiative) counts for each of them. The count also includes bills rejected in the preliminary reading, not only bills that became law - this is a number about initiative, not about legislative success.

05Topic tagging

Votes are tagged by topic (the draft law, the cost of living, etc.) manually or by keywords, and not every vote is tagged - a large share is tagged "Other" (mostly narrow or procedural proposals). The number of votes on a given topic may look slightly different between views on the site (for example on the topic page versus an MK's votes tab) if one includes procedural votes and the other does not - this is a difference in scope, not a contradiction in the data.

06Party position by topic

The position shown (for/against/mixed) is a simple majority of the actual votes of the party's members on the bills tagged to that topic - not a policy statement by the party. Under the label, the exact breakdown is always shown (how many bills were decided, how many of them were initiated by the coalition and how many by the opposition), because the coalition or opposition majority tends to vote as a bloc on bills that are not its own - which can turn an inherently neutral topic (such as the cost of living) into a label that looks political, when in practice the vote mainly reflected who initiated the bill, not a position on the merits. See also Corrections about a bug that caused some rows to show "0 bills".

07The quiz "Which list is closest to you?"

On each question, the match is 100% when you and the list are at the same point on the axis, 0% when you are at opposite ends, and linear in between. The overall percentage is the average over the questions you answered on which the list has a documented position. So that a list close to you on many questions ranks above one close to you on few, every question you answered on which the list has no documented position counts as half a question with a 50% match. A list with a documented position on fewer than half of the questions you answered is shown separately, because a match on few questions is not a result worth relying on. "In the middle" counts as the midpoint of the axis; if you are not sure, you skip the question and it is left out of the calculation. The answers and the closest list are saved anonymously, with no name, contact details or IP address, and are used only for general statistics (see the privacy policy). The result is a comparison, not a recommendation.

08The "Whose tweet is this?" game

A selection of real tweets by sitting MKs from the current term. Tweets that are retweets ("RT @...", not the member's own words) and tweets that mention the author's full name (which would give the answer away) are removed from the pool in advance. Wrong answers (distractors) are chosen from MKs of the same gender and the same coalition/opposition side where possible, so that guessing by grammar (a verb in the masculine or feminine form) or by political stereotype is not enough on its own.

09Legal proceedings

Only proceedings documented in a reliable public source are shown, in neutral wording and under the presumption of innocence, with a link to the source. The list is checked against all sitting MKs, using survey sources (media overviews of criminal proceedings against MKs), not by searching name after name.

On the distribution

In the latest scan, two major independent media overviews of MKs in criminal proceedings listed only coalition members. That is, the concentration in the list mainly reflects a fact about the period, not the way we collect. Even so, the scan also added a case from the opposition that had been missing, and dropped a coalition case whose file was closed without findings.

10Ministerial committees

Unlike Knesset committees (pulled directly from Knesset data), membership of government committees is not published in a structured database, and it changes with every government decision. The site shows only the two committees that have a fixed, known membership rule: the Ministerial Committee for National Security Affairs (membership fixed by law) and the Ministerial Committee for Legislation (membership fixed by long-standing custom - the Justice Minister is always chair, the Finance Minister always a member). The rule is applied to whoever currently holds the relevant office according to the site's own data - not to additional, changing appointments to these committees or to other government committees, which cannot be reliably verified against the government membership shown on the site.

11Data updates

The data is synced automatically from the official sources on an ongoing basis. The date of the last update of the MKs' data appears at the bottom of every page on the site.