
We develop a model of a sealed-bid first-price auction where the auctioneer, acting on behalf of the seller, may collude with one of the bidders. This bidder can offer a bribe to the auctioneer to see the opponents bids before making his own bid. We show that the seller often wants to offer the auctioneer a remuneration scheme that depends on the collected revenue and prevents the auctioneer from accepting the bribe even if the legal anti-corruption institutions are weak. We consider both cases of bidders who are informed and bidders who are uninformed about corruption.
The role of digitalization is becoming increasingly important. By developing Python codes, we have obtained annual report data from Chinese A-share listed companies, enabling us to quantify their product quality. Empirically, we have discovered a significant U-shaped relationship between the extent of digitalization of enterprises and the quality of their products. Furthermore, mechanism analysis reveals that digitalization effectively enhances the specialization level of these listed companies, thereby leading to an improvement in product quality. Moreover, this paper delves into the heterogeneity of the impact of digitalization on product quality. Based on the findings of this paper, we provide some policy implications.
This paper models pollution with emissions both in the production and consumption of goods. Modeling simultaneous consumption and production pollution, and the effectiveness of import and export taxes across international jurisdictions are contributions to the literature. Two duopoly games are investigated with polluting firms/consumers located in different countries. Results show that for given import/export taxes, trade liberalization reduces pollution emissions, consumer surplus, and producer surplus. Further, Lerners symmetry theorem holds with production and consumption pollution. Export taxes generate greater welfare and less social damage than import taxes, and countries choose a closed economy with endogenous import/export taxes.
The dominant perspective in the literature exploring the relationship between regulation and societal trust suggests that there is a negative correlation between the thickness of regulation and the prevailing levels of trust in a society (see Aghion et al., 2010). Our paper highlights instead that a positive complementarity exists under plausible circumstances between trust, the enforcement of regulation, and corporate social responsibility (CSR) or environmental and social (ES) practices. Using a relational contracting approach, we show that regulatory enforcement reduces the misalignment of incentives between firms and consumers, making it easier for trust and CSR to arise endogenously.
This paper first surveys a variety of problems endemic in the notion of ordinary meaning and then argues in favor of a semantic approach to ordinary meaning that is intensionalist, prototype-based, and corpus-linguistic in nature. This approach uses embeddings models rather than large language models (LLMs), which makes it more replicable, versatile (it can be applied to words that did not exist at corpus time), and more cheaply/quickly adoptable than the current traditional practice. I exemplify the approach using two embeddings models - one from a 2014 Common Crawl of the WWW, the other trained on 1950s American English corpus data.
This article reexamines Martin Bubers writings from the First World War, arguing that they constitute a philosophically coherent expression of Lebensphilosophie, rather than a transient deviation preceding I and Thou. Against the prevailing tendency to minimize or bracket these texts, it situates Bubers wartime thought within the broader philosophy of life constellation shaped by Simmel, Bergson, Nietzsche and Dilthey. It shows how Buber interprets the war through kinesis, Erlebnis, temporal rupture and organic unity, reading it as a regenerative moment that discloses lifes creative and unconditioned forces. The article further demon-strates that Buber radicalizes Lebensphilosophie by integrating it into a specifically Jewish vision of communal renewal and historical vocation. Contextualizing these reflections reveals how this phase exposes both the generative scope and internal dynamics of Lebensphilosophie, while also inviting a reconsideration of Bubers later dialogical turn as an ethical reconfiguration of the philosophy of life within the interpersonal sphere of relation.
Legal judgment is threatened by a plethora of biases. Among those, the effect of outcome bias-the tendency to let outcome severity distort assessments of mental states and culpability-is particularly well documented and its impact is hypothesized to be pervasive. AI advisory systems could in principle help alleviate human bias, but Almeida et al. (2024) found substantial outcome effects in several commercial LLMs, suggesting AI judgment may be comparably biased. We reassess these findings using eight current open-source models (Cogito, Deepseek-R1, Gemma 3, Llama 3.1, Mistral, Phi 4) across multiple parameter sizes and temperature settings, replicating the Flood Paradigm from Kneer and Skoczeń (2023). Using a multilevel model with 100 responses per condition per temperature, we find that current open-source models exhibit little to no outcome bias-in stark contrast to both human participants and the older LLMs tested previously. Only Cogito and Deepseek-R1 show small effects, and temperature variation contributes negligibly. However, models differ substantially in baseline judgment severity-Mistral rates culpability considerably higher than Gemma 3 or Phi 4demonstrating that absence of outcome bias does not guarantee neutrality. Our findings support cautious optimism about research into open-source LLMs as debiasing tools in legal contexts, while emphasizing that bias is multidimensional and responsible deployment requires evaluation across bias types and doctrinal domains.
Before indulging in the perplexing prospect of legal reasoning by machines, we must recognize that machine learning is built upon, closely tied with, and compared against human performance on this task. Yet the predominant strandof American judicial reasoning violates norms of principled legal reasoning. From a psychological perspective, these violations can be understood as the natural offshoots of the cognitive processes that make complex decision making possible. These violations have potential implications for legal reasoning by Artificial Intelligence as judicial opinions are an ideal source of training material for AI, especially in Common Law regimes. The question is whether we should prompt machines to imitate human judges or to strip away the imprints of human foibles and biases that judges imprint on their opionions. The latter option opens the door to the prospect of machines showing greater fidelity to the law than the judges in whom we entrust this vital societal function.
The precision of legal commands - how finely customized are they to the specific circumstances - could be dramatically increased using AI algorithms trained to identify factors relevant to the optimal tailoring of any legal rule. To accomplish such automation of the law, a legal rule must be enacted with explicit statement of its objectives and their relative weight, instructing algorithms what to optimize. This paper examines whether such upfront quantification of a rules objectives is possible and even desirable.
While statistical machine learning has advanced legal AI, its reliance on probabilities conflicts with some of the legal systems needs. This position paper argues that the strength of current AI tools, chiefly based on statistical learning, is also their main weakness in the legal domain. Instead, narrative-based approaches, inspired by Conviction Narrative Theory (CNT) and Algorithmic Information Theory (AIT), offer a better alternative. CNT explains how humans construct explanations and make decisions through coherent narratives, while AIT can be used to quantify the plausibility of these narratives. The paper discusses the applications of both paradigms to the legal domain and to legal AI, and in particular, how legal doctrine fits within CNT. Preliminary experiments on LLMs conclude this paper.
Language models (LMs) have shown outstanding performance in legal text summarization. It is crucial that personally identifying information (PII) included in the source document should not leak in the summary. Prior efforts have mostly focused on studying how LMs may inadvertently leak PII from training data. However, to what extent LMs can provide privacy-preserving summaries given a non-private source legal document remains under-explored. In this paper, we perform a preliminary empirical study on privacy-preservation in legal text summarization showing that LMs often cannot prevent PII leakage in summaries.
Despite generative AIs growing technical capacity to transform legal work, legal practice has changed little. This essay explores a cognitive explanation: lawyers often struggle to deeply understand AI-generated responses to com-plex or unfamiliar legal issues. These challenges impair their ability to evaluate AI outputs and perform related tasks such as client advising, argument tailoring, or issue synthesis. The problem is especially acute for junior lawyers and may reflect core features of legal reasoning. These findings highlight persistent cognitive constraints on professional adaptation and raise broader questions about human-AI interaction in expert domains.
This Comment shows how large language models (LLMs) can help courts discern the ordinary meaning of statutory terms. Instead of relying on expert-heavycorpus‑linguistic techniques (Gries, 2026), the author simulates a human survey with GPT‑4o. Demographically realistic AI agents replicate the 2,835 participants in Tobias 2020 study on vehicle and yield response distributions with no statistically significant difference from the human data (Kolmogorov-Smirnov p = 0.915). The paper addresses concerns about hallucinations, reproducibility, training-data contamination, and explainability, and introduces the locked‑prompt Ordinary Meaning Bot, arguing that LLM-based survey simulation is a practical, accurate alternative to dictionaries, intuition, or complex corpus analysis.