Skip to content

Latest commit

 

History

History
150 lines (127 loc) · 6.49 KB

basic_interface.rst

File metadata and controls

150 lines (127 loc) · 6.49 KB
Raw

Basic SP Network Interface Proposal

I propose that the review site would present each paper with a basic tabbed interface, initially with just two tabs, Recommendations and Discussion. The main difference is that whereas recommendations are automatically forwarded to the recommender's subscribers, discussion comments are not; they are simply displayed on the Discussion tab for the paper. The Recommendations tab would show who recommended the paper, for what area(s), and what they said about it. The Discussions tab would show the usual threaded discussion interface that everyone is used to.

The Recommendations tab

This would show the following simple interface that lets you view and make recommendations. It should be designed to look like the arXiv main page for a paper (e.g. sidebox for accessing PDF, PostScript etc. for the paper), but with the main content of the page shown in the following order (see explanatory notes below on the interactive elements):

  • paper title, authors
  • [+] abstract
  • "Flagging papers that interest you will help this system predict other papers that may interest you. Subscribing to people who recommended this paper will alert you to other papers they recommend. Writing recommendations that other people find useful will attract them to subscribe to receive your future recommendations and papers."
  • [STAR] Rated must-read by John Baez, Ian Fleming and 37 others [+]
  • If this paper interests you, flag the group(s) you would recommend read it:

[ ] logtree phylogeny: rec'd by John Baez, Kevin Costner and 21 others [+] [ ] ultimate wrestling: rec'd by Herbert Spencer... [+] ... (more topics) ... Add a topic-group: _____________

  • News & Views
    • [+] John Baez: The most exciting paper since...
    • [+] Ian Fleming: A license to kill...
    • [+] 5 others
    • Tell the world! Write a News & Views:

Interface Notes

  • clicking +/- sign shows or hides the text of the abstract.
  • clicking the STAR icon toggles whether you rate the paper must-read or not; clicking the +/- sign toggles viewing of the list of other recommenders.
  • the list of candidate topic-groups would be the superset of what other people (including the authors) suggested as groups that would be interested in this paper, and the groups the user belongs to.
  • Add a topic group would let the user enter text to search the database of existing topic-groups and pick one or alternatively create a completely new group.
  • "News & Views" is just a provisional title, suggestive of alerting others to something new and interesting and of communicating your opinion about what matters. "News & Views" is what Nature calls these kind of highlight pieces.
  • A News & Views item is implemented essentially a discussion item that gets forwarded to its authors subscribers, who in turn can recommend it to their subscribers, and so on. (in this sense it is like a publication in its own right, but of course we won't allow people to write News & Views on a News & Views item!). As a discussion item, it can be commented on by others just like any discussion item).
  • expanding a News & Views will show its text, as well as a link for viewing comments / commenting on it in the Discussion tab.
  • you can add a News & Views by giving a DOI or URL for something published elsewhere (e.g. on your blog), or by simply entering text into a textbox here.

The Discussion tab

This would be a generic threaded discussion for this paper. As long as people can add new topics for discussion and comment on existing topics, it will be adequate for an initial release. Over time it would add peer review features (e.g. you could flag a discussion item as raising doubt about the validity of a major claim in the paper). However that is not needed at first.

Validation

Standard peer review requires validation before publication. We could incorporate a similar "gateway" requirement as follows:

  • at least one referee would have to certify the paper's results as a significant advance over previous work, before any recommendations (made by that referee or anyone else) would be sent to subscribers.
  • in the absence of any such validation, recommendations would be visible on the homepage for the paper (but would not be sent to subscribers).
  • if you recommend a paper that hasn't been validated, you'll get a message like "No one has validated this paper's results yet. Recommendations will not be sent until you (or someone else) marks its results as a valid, significant advance over previous work. Would you like to proceed to the validation page?"

Validation would consist of two basic pieces:

  • listing the basic questions the paper addresses, and its basic claims (answering those questions).
  • "Validation statement":
    • you have read the full paper;
    • you have listed all of your major concerns about the paper's claims, and your level of confidence;
    • based on this, you consider the paper a valid, significant advance over previous literature.

User registration

  • users would be asked to register with email address that matches their arXiv account, and would then receive an email for activating their account. This way we link each user to their papers etc. in arXiv.
  • to seed the system for recommending papers to them, new users would be asked to list recent papers they considered must-read for their own work, and / or other researchers whose interests they consider to be most similar to their own.
  • users would be asked to select topic-groups that fit their research interests. Initial topics can be suggested by the system based on the papers and people they listed, e.g. if any of those people already listed some topic-groups, offer those as options. Of if any other people listed those same papers or people, offer the topic-groups that those people listed. Of course the user can add his own topic-group terms.
  • the site would follow the Amazon model of remembering the user (i.e. based on their last login to this site on this computer, show their personal recommendations), but authenticate them (i.e. ask for password) if they want to publish a recommendation.

Look & Feel

I suggest the look & feel of the site draw from two main sources: