Smart Broadcasting

Distilling Information Reliability and Source Trustworthiness from Digital Traces

“…However, previous work typically shares one or more of the following limitations, which we address in this work: (i) they only support knowledge triplets (subject, predicate, object) or structured knowledge; (ii) they assume there is a truth, however, a statement may be under discussion when a source writes about it; and, (iii) they do not distinguish between the unreliability of the knowledge item to which the statement refers and the trustworthiness of the source. Temporal point processes have been previously used to model information cascades [15,11,5], social activity [14,19,12], badges [10], network evolution [18,13], opinion dynamics [6], or product competition [26]. However, to the best of our knowledge, the present work is the first that leverages temporal point processes in the context of information reliability and source trustworthiness.…”

Deep Reinforcement Learning of Marked Temporal Point Processes

“…In the remainder of the paper, whenever an intensity function and mark distribution are parametrized by θ, we write λ * θ (•), m * θ (•), P θ (A T ), and, for notational simplicity, use p * θ = (λ * θ , m * θ ) as a short-hand to denote the joint probability density of the MTPP. Recent literature [5,8,12,13,17,30,33] has established that MTPPs outperform other models (e.g., exponential law) in their ability to accurately predict online and off-line human actions.…”

“…Following previous work [29,33,34], we measure visibility a user achieves, i.e., the reward, using two different metrics: (i) the position of her most recent post on her followers' feeds over time, or rank, i.e., R * (T ) = T 0 r(t)dt, where the position zero, r(t) = 0, corresponds to top and thus lower is better; (ii) the (amount of) time that her most recent post is at the top of her followers' feeds, or time at the top, i.e., R * (T ) = T 0 I(r(t) < 1)dt, and thus higher is better. If the followers' feeds are sorted in reverse chronological order, previous work has derived optimal offline [12] and online [34] algorithms for (i) and (ii), respectively, under the additional assumption that the posting intensity of other users her followers follow adopts certain functional form. However, as pointed out by previous work, feeds are typically algorithmically sorted, the posting intensity of other users may be highly complex, and thus the derived algorithms may be of limited use in practice.…”

“…Since we do not have access to the feed sorting algorithm used by Twitter, we experiment with a relatively simple sorting algorithm based on a priority queue 6 (refer to Appendix F). Here, since our feed sorting algorithm does only depends on the time of the post and the identity of (RQ), a variant of RedQueen which has access to true ranks (RQ * ), and Karimi's method [12] on feeds using a sorting algorithm based on a priority queue (refer to Appendix F). Panels (a) and (b) show the average rank and time at the top, where the solid horizontal line shows the median value across users, normalized with respect to the value achieved by a user who follows a uniform Poisson intensity, and the box limits correspond to the 25%-75% percentiles.…”

“…In contrast, previous works considered the policy to be a probability distribution or, more rarely, a deterministic function [4,9,28]. Finally, we apply our methodology to two different applications in personalized teaching [14,22,27] and viral marketing [12,25,29,33,34], respectively. For simple dynamics and objective functions, which allow for stochastic optimal control approaches, our method achieves a comparable performance even though it does not have access to the true underlying dynamics.…”