The release of openly available, robust natural language generation algorithms (NLG) has spurred much public attention and debate. One reason lies in the algorithms' purported ability to generate human-like text across various domains. Empirical evidence using incentivized tasks to assess whether people (a) can distinguish and (b) prefer algorithm-generated versus human-written text is lacking. We conducted two experiments assessing behavioral reactions to the state-of-the-art Natural Language Generation algorithm GPT-2 (Ntotal = 830). Using the identical starting lines of human poems, GPT-2 produced samples of poems. From these samples, either a random poem was chosen (Human-out-of-the-loop) or the best one was selected (Human-in-the-loop) and in turn matched with a human-written poem. In a new incentivized version of the Turing Test, participants failed to reliably detect the algorithmically-generated poems in the Human-in-the-loop treatment, yet succeeded in the Human-out-of-the-loop treatment. Further, people reveal a slight aversion to algorithm-generated poetry, independent on whether participants were informed about the algorithmic origin of the poem (Transparency) or not (Opacity). We discuss what these results convey about the performance of NLG algorithms to produce human-like text and propose methodologies to study such learning algorithms in human-agent experimental settings.

We use the theory of cooperative games for the design of fair insurance contracts. An insurance contract needs to specify the premium to be paid and a possible participation in the benefit (or surplus) of the company. It results from the analysis that when a contract is exposed to the default risk of the insurance company, ex-ante equilibrium considerations require a certain participation in the benefit of the company to be specified in the contracts. The fair benefit participation of agents appears as an outcome of a game involving the residual risks induced by the default possibility and using fuzzy coalitions.

Up-to-date poverty maps are an important tool for policymakers, but until now, have been prohibitively expensive to produce. We propose a generalizable prediction methodology to produce poverty maps at the village level using geospatial data and machine learning algorithms. We tested the proposed method for 25 Sub-Saharan African countries and validated them against survey data. The proposed method can increase the validity of both single country and cross-country estimations leading to higher precision in poverty maps of the 25 countries than previously available. More importantly, our cross-country estimation enables the creation of poverty maps when it is not practical or cost-effective to field new national household surveys, as is the case with many Sub-Saharan African countries and other low- and middle-income countries.

Much of the research on networks using patent data focuses on citations and the collaboration networks of inventors, hence regarding patents as a positive sign of invention. However, patenting is, most importantly, a strategic action used by companies to compete with each other. This study sheds light on inter-organisational adversarial relationships in patenting for the first time. We constructed and analysed the network of companies connected via patent opposition relationships that occurred between 1980 and 2018. A majority of the companies are directly or indirectly connected to each other and hence form the largest connected component. We found that in the network, many companies disapprove patents in various industrial sectors as well as those owned by foreign companies. The network exhibits heavy-tailed, power-law-like degree distribution and assortative mixing, making it an unusual type of topology. We further investigated the dynamics of the formation of this network by conducting a temporal network motif analysis, with patent co-ownership among the companies considered. By regarding opposition as a negative relationship and patent co-ownership as a positive relationship, we analysed where collaboration may occur in the opposition network and how such positive relationships would interact with negative relationships. The results identified the structurally imbalanced triadic motifs and the temporal patterns of the occurrence of triads formed by a mixture of positive and negative relationships. Our findings suggest that the mechanisms of the emergence of the inter-organisational adversarial relationships may differ from those of other types of negative relationships hence necessitating further research.

We discuss when law-invariant convex functionals "collapse to the mean". More precisely, we show that, in a large class of spaces of random variables and under mild semicontinuity assumptions, the expectation functional is, up to an affine transformation, the only law-invariant convex functional that is linear along the direction of a nonconstant random variable with nonzero expectation. This extends results obtained in the literature in a bounded setting and under additional assumptions on the functionals. We illustrate the implications of our general results for pricing rules and risk measures.

We establish a variety of numerical representations of preference relations induced by set-valued risk measures. Because of the general incompleteness of such preferences, we have to deal with multi-utility representations. We look for representations that are both parsimonious (the family of representing functionals is indexed by a tractable set of parameters) and well behaved (the representing functionals satisfy nice regularity properties with respect to the structure of the underlying space of alternatives). The key to our results is a general dual representation of set-valued risk measures that unifies the existing dual representations in the literature and highlights their link with duality results for scalar risk measures.

This research analyses high-frequency data of the cryptocurrency market in regards to intraday trading patterns related to algorithmic trading and its impact on the European cryptocurrency market. We study trading quantitatives such as returns, traded volumes, volatility periodicity, and provide summary statistics of return correlations to CRIX (CRyptocurrency IndeX), as well as respective overall high-frequency based market statistics with respect to temporal aspects. Our results provide mandatory insight into a market, where the grand scale employment of automated trading algorithms and the extremely rapid execution of trades might seem to be a standard based on media reports. Our findings on intraday momentum of trading patterns lead to a new quantitative view on approaching the predictability of economic value in this new digital market.

A simulator of individual claim experience called SynthETIC is described. It is publicly available, open source and fills a gap in the non-life actuarial toolkit. It simulates, for each claim, occurrence, notification, the timing and magnitude of individual partial payments, and closure. Inflation, including (optionally) superimposed inflation, is incorporated in payments. Superimposed inflation may occur over calendar or accident periods. The claim data are summarized by accident and payment "periods" whose duration is an arbitrary choice (e.g. month, quarter, etc.) available to the user. The code is structured as eight modules (occurrence, notification, etc.), any one or more of which may be re-designed according to the user's requirements. The default version is parameterized so as to include a broad (though not numerically precise) resemblance to the major features of experience of a specific (but anonymous) Auto Bodily Injury portfolio. It thus reflects a number of desirable (but complicated) data features, but the general structure is suitable for most lines of business, with some amendment of modules. The structure of the simulator enables the inclusion of a number of important dependencies between the variables related to an individual claim, e.g. dependence of notification delay on claim size, of the size of a partial payment on the sizes of those preceding, etc. The user has full control of the mechanics of the evolution of an individual claim. As a result, the complexity of the data set generated (meaning the level of difficulty of analysis) may be dialled anywhere from extremely simple to extremely complex. The user may generate a collection of data sets that provide a spectrum of complexity, and the collection may be used to present a model under test with a steadily increasing challenge.

This paper undertakes a near real-time analysis of the income distribution effects of the COVID-19 crisis in Australia to understand the ongoing changes in the income distribution as well as the impact of policy responses. By semi-parametrically combining incomplete observed data from three different sources, namely, the Monthly Longitudinal Labour Force Survey, the Survey of Income and Housing and the administrative payroll data, we estimate the impact of COVID-19 and the associated policy responses on the Australian income distribution between February and June 2020, covering the immediate periods before and after the initial outbreak. Our results suggest that despite the growth in unemployment, the Gini of the equalised disposable income inequality dropped by nearly 0.03 point since February. The reduction is because of the additional wage subsidies and welfare supports offered as part of the policy response, offsetting a potential surge in income inequality. Additionally, the poverty rate, which could have been doubled in the absence of the government response, also reduced by 3 to 4 percentage points. The result shows the effectiveness of temporary policy measures in maintaining both the living standards and the level of income inequality. However, the heavy reliance on the support measures raises the possibility that the changes in the income distribution may be reversed and even substantially worsened off should the measures be withdrawn.

In this paper we study a class of time-inconsistent terminal Markovian control problems in discrete time subject to model uncertainty. We combine the concept of the sub-game perfect strategies with the adaptive robust stochastic to tackle the theoretical aspects of the considered stochastic control problem. Consequently, as an important application of the theoretical results, by applying a machine learning algorithm we solve numerically the mean-variance portfolio selection problem under the model uncertainty.

We discuss the impact of a Covid-like shock on a simple toy economy, described by the Mark-0 Agent-Based Model that we developed and discussed in a series of previous papers. We consider a mixed supply and demand shock, and show that depending on the shock parameters (amplitude and duration), our toy economy can display V-shaped, U-shaped or W-shaped recoveries, and even an L-shaped output curve with permanent output loss. This is due to the existence of a self-sustained "bad" state of the economy. We then discuss two policies that attempt to moderate the impact of the shock: giving easy credit to firms, and the so-called helicopter money, i.e. injecting new money into the households savings. We find that both policies are effective if strong enough, and we highlight the potential danger of terminating these policies too early. While we only discuss a limited number of scenarios, our model is flexible and versatile enough to allow for a much wider exploration, thus serving as a useful tool for the qualitative understanding of post-Covid recovery. We provide an on-line version of the code at https://gitlab.com/sharma.dhruv/markovid .