summaryrefslogtreecommitdiffstats
path: root/project/main.tex
diff options
context:
space:
mode:
Diffstat (limited to 'project/main.tex')
-rw-r--r--project/main.tex24
1 files changed, 19 insertions, 5 deletions
diff --git a/project/main.tex b/project/main.tex
index 34a37f7..a0c58a6 100644
--- a/project/main.tex
+++ b/project/main.tex
@@ -71,11 +71,25 @@ that we are planning to consider:
end of the auction, but also maintains a distribution on the
probability of being allocated.
\item at each time step, the bidders update their distribution based on
- what has been observed at the previous step (hence the bidders'
- distribution are a function of the past history up to this point)
- \item at each time step, bidders place in order to maximize a function
- which is increasing in the probability of being allocated and their
- utility: contrary to
+ what has been observed at the previous step. Hence the bidders'
+ distribution are a function of the past history up to this point.
+ \item at each time step, if a bidder is not currently winning, she is
+ willing to sacrifice some expected utility to increase their
+ probability of being allocated. That is, the agents' are not fully
+ rational with respect to their ``outside-the-auction utility'': while
+ the auction is still running, the utility function that they are
+ maximizing puts a higher weight on simply winning the auction.
+\end{itemize}
+
+Under these behavioral axioms, we want to analyze the eBay-style auctions with
+respect to the following properties: truthfulness, envy-freeness, revenue
+optimality. One last property that we would like to analyze is \emph{regret}:
+we don't know yet how to define it properly, but a good notion of regret should
+be a function of the difference in utility induced by the ``non-rational''
+behavior of the agents: during the auction they are maximizing a different
+utility function (biased by the probability of winning); this might decrease
+their true utility.
+
\end{itemize}
\section{Relevant Prior Work} We are planning to consider prior work on first-price auctions where the bidders have multiple rounds of bidding over time, and various mechanisms that can capture this idea. One inspiration comes from the \lq\lq bid-your-utility \rq\rq auction described in the following paper: Hoy, D., Jain, K., \& Wilkens, C. A. (2013, June). A dynamic axiomatic approach to first-price auctions. In Proceedings of the fourteenth ACM conference on Electronic commerce (pp. 583-584). ACM.