Using the Veil of Ignorance to align AI systems with principles of justice.

The philosopher John Rawls proposed the Veil of Ignorance (VoI) as a thought experiment to identify fair principles for governing a society. Here, we apply the VoI to an important governance domain: artificial intelligence (AI). In five incentive-compatible studies (N = 2, 508), including two preregistered protocols, participants choose principles to govern an Artificial Intelligence (AI) assistant from behind the veil: that is, without knowledge of their own relative position in the group. Compared to participants who have this information, we find a consistent preference for a principle that instructs the AI assistant to prioritize the worst-off. Neither risk attitudes nor political preferences adequately explain these choices. Instead, they appear to be driven by elevated concerns about fairness: Without prompting, participants who reason behind the VoI more frequently explain their choice in terms of fairness, compared to those in the Control condition. Moreover, we find initial support for the ability of the VoI to elicit more robust preferences: In the studies presented here, the VoI increases the likelihood of participants continuing to endorse their initial choice in a subsequent round where they know how they will be affected by the AI intervention and have a self-interested motivation to change their mind. These results emerge in both a descriptive and an immersive game. Our findings suggest that the VoI may be a suitable mechanism for selecting distributive principles to govern AI.

Spurious normativity enhances learning of compliance and enforcement behavior in artificial agents.

How do societies learn and maintain social norms? Here we use multiagent reinforcement learning to investigate the learning dynamics of enforcement and compliance behaviors. Artificial agents populate a foraging environment and need to learn to avoid a poisonous berry. Agents learn to avoid eating poisonous berries better when doing so is taboo, meaning the behavior is punished by other agents. The taboo helps overcome a credit assignment problem in discovering delayed health effects. Critically, introducing an additional taboo, which results in punishment for eating a harmless berry, further improves overall returns. This "silly rule" counterintuitively has a positive effect because it gives agents more practice in learning rule enforcement. By probing what individual agents have learned, we demonstrate that normative behavior relies on a sequence of learned skills. Learning rule compliance builds upon prior learning of rule enforcement by other agents. Our results highlight the benefit of employing a multiagent reinforcement learning computational model focused on learning to implement complex actions.

Impacts of different land management histories on forest change.

Many western North American forest types have experienced considerable changes in ecosystem structure, composition, and function as a result of both fire exclusion and timber harvesting. These two influences co-occurred over a large portion of dry forests, making it difficult to know the strength of either one on its own or the potential for an interaction between the two. In this study, we used contemporary remeasurements of a systematic historical forest inventory to investigate forest change in the Sierra Nevada. The historical data opportunistically spanned a significant land management agency boundary, which protected part of the inventory area from timber harvesting. This allowed for a robust comparison of forest change between logged and unlogged areas. In addition, we assessed the effects of recent management activities aimed at forest restoration relative to the same areas historically, and to other areas without recent management. Based on analyses of 22,007 trees (historical, 9,573; contemporary, 12,434), live basal area and tree density significantly increased from 1911 to the early 2000s in both logged and unlogged areas. Both shrub cover and the proportion of live basal area occupied by pine species declined from 1911 to the early 2000s in both areas, but statistical significance was inconsistent. The most notable difference between logged and unlogged areas was in the density of large trees, which declined significantly in logged areas, but was unchanged in unlogged areas. Recent management activities had a varied impact on the forest structure and composition variables analyzed. In general, areas with no recent management activities experienced the greatest change from 1911 to the early 2000s. If approximating historical forest conditions is a land management goal the documented changes in forest structure and composition from 1911 to the early 2000s indicate that active restoration, including fire use and mechanical thinning, is needed in many areas.

Mini-review: Assessing the drivers of ship biofouling management--aligning industry and biosecurity goals.

Biofouling exerts a frictional and cost penalty on ships and is a direct cause of invasion by marine species. These negative consequences provide a unifying purpose for the maritime industry and biosecurity managers to prevent biofouling accumulation and transfer, but important gaps exist between these sectors. This mini-review examines the approach to assessments of ship biofouling among sectors (industry, biosecurity and marine science) and the implications for existing and emerging management of biofouling. The primary distinctions between industry and biosecurity in assessment of vessels biofouling revolve around the resolution of biological information collected and the specific wetted surface areas of primary concern to each sector. The morphological characteristics of biofouling and their effects on propulsion dynamics are of primary concern to industry, with an almost exclusive focus on the vertical sides and flat bottom of hulls and an emphasis on antifouling and operational performance. In contrast, the identity, biogeography, and ecology of translocated organisms is of highest concern to invasion researchers and biosecurity managers and policymakers, especially as it relates to species with known histories of invasion elsewhere. Current management practices often provide adequate, although not complete, provision for hull surfaces, but niche areas are well known to enhance biosecurity risk. As regulations to prevent invasions emerge in this arena, there is a growing opportunity for industry, biosecurity and academic stakeholders to collaborate and harmonize efforts to assess and manage biofouling of ships that should lead to more comprehensive biofouling solutions that promote industry goals while reducing biosecurity risk and greenhouse gas emissions.

Novel characterization of landscape-level variability in historical vegetation structure.

We analyzed historical timber inventory data collected systematically across a large mixed-conifer-dominated landscape to gain insight into the interaction between disturbances and vegetation structure and composition prior to 20th century land management practices. Using records from over 20 000 trees, we quantified historical vegetation structure and composition for nine distinct vegetation groups. Our findings highlight some key aspects of forest structure under an intact disturbance regime: (1) forests were low density, with mean live basal area and tree density ranging from 8-30 m2 /ha and 25-79 trees/ha, respectively; (2) understory and overstory structure and composition varied considerably across the landscape; and (3) elevational gradients largely explained variability in forest structure over the landscape. Furthermore, the presence of large trees across most of the surveyed area suggests that extensive stand-replacing disturbances were rare in these forests. The vegetation structure and composition characteristics we quantified, along with evidence of largely elevational control on these characteristics, can provide guidance for restoration efforts in similar forests.

Growth and development of the barnacle Amphibalanus amphitrite: time and spatially resolved structure and chemistry of the base plate.

The radial growth and advancement of the adhesive interface to the substratum of many species of acorn barnacles occurs underwater and beneath an opaque, calcified shell. Here, the time-dependent growth processes involving various autofluorescent materials within the interface of live barnacles are imaged for the first time using 3D time-lapse confocal microscopy. Key features of the interface development in the striped barnacle, Amphibalanus (= Balanus) amphitrite were resolved in situ and include advancement of the barnacle/substratum interface, epicuticle membrane development, protein secretion, and calcification. Microscopic and spectroscopic techniques provide ex situ material identification of regions imaged by confocal microscopy. In situ and ex situ analysis of the interface support the hypothesis that barnacle interface development is a complex process coupling sequential, timed secretory events and morphological changes. This results in a multi-layered interface that concomitantly fulfills the roles of strongly adhering to a substratum while permitting continuous molting and radial growth at the periphery.

Scaffolding cooperation in human groups with deep reinforcement learning.

Effective approaches to encouraging group cooperation are still an open challenge. Here we apply recent advances in deep learning to structure networks of human participants playing a group cooperation game. We leverage deep reinforcement learning and simulation methods to train a 'social planner' capable of making recommendations to create or break connections between group members. The strategy that it develops succeeds at encouraging pro-sociality in networks of human participants (N = 208 participants in 13 groups) playing for real monetary stakes. Under the social planner, groups finished the game with an average cooperation rate of 77.7%, compared with 42.8% in static networks (N = 176 in 11 groups). In contrast to prior strategies that separate defectors from cooperators (tested here with N = 384 in 24 groups), the social planner learns to take a conciliatory approach to defectors, encouraging them to act pro-socially by moving them to small highly cooperative neighbourhoods.

Learning few-shot imitation as cultural transmission.

Cultural transmission is the domain-general social skill that allows agents to acquire and use information from each other in real-time with high fidelity and recall. It can be thought of as the process that perpetuates fit variants in cultural evolution. In humans, cultural evolution has led to the accumulation and refinement of skills, tools and knowledge across generations. We provide a method for generating cultural transmission in artificially intelligent agents, in the form of few-shot imitation. Our agents succeed at real-time imitation of a human in novel contexts without using any pre-collected human data. We identify a surprisingly simple set of ingredients sufficient for generating cultural transmission and develop an evaluation methodology for rigorously assessing it. This paves the way for cultural evolution to play an algorithmic role in the development of artificial general intelligence.

Considering Commercial Vessels as Potential Vectors of Stony Coral Tissue Loss Disease.

Stony coral tissue loss disease (SCTLD) is a troubling new disease that is spreading rapidly across the greater Caribbean region, but the etiological agent(s) and the mechanisms(s) of spread are both unknown. First detected off the coast of Miami, Florida, major ocean currents alone do not explain the pattern of spread, with outbreaks occurring across geographically disjunct and distant locations. This has raised concerns by researchers and resource managers that commercial vessels may contribute as vectors to spread of the disease. Despite existing regulatory and management strategies intended to limit coastal marine invasion risks, the efficacy of these measures is still unresolved for ship-borne microorganisms, and disease transport via ballast water and hull biofouling are under examination given the high ship traffic in the region. Here, to help inform the discussion of ships as possible vectors of SCTLD, we provide an overview of the current state of knowledge about ships and their potential to transfer organisms in the greater Caribbean, focusing in particular on ballast water, and outline a set of recommendations for future research.

Design and installation of ballast water sample ports: Current status and implications for assessing compliance with discharge standards.

To verify ships' compliance with ballast water regulations, samples may be collected and tested for viable organisms. This task is completed using a sample probe, which is placed in the ballast discharge pipe through a sample port (a flanged opening). To collect representative samples, the placement of the sample port and the size of the sample probe must be appropriate for the shipboard piping arrangement and ballast water flows. The placement of sample ports was evaluated on 72 ships to assess the current condition of ballast water sampling installations against available guidance. Few ships (15%) had sample ports fully aligned with International Organization for Standardization (ISO) standard 11711-1. While current configurations may present challenges in collecting representative samples, these installations likely occurred before the ISO standard was available. Future installations should be in accordance with the standard to facilitate representative sampling.

Barnacle cement: a polymerization model based on evolutionary concepts.

Enzymes and biochemical mechanisms essential to survival are under extreme selective pressure and are highly conserved through evolutionary time. We applied this evolutionary concept to barnacle cement polymerization, a process critical to barnacle fitness that involves aggregation and cross-linking of proteins. The biochemical mechanisms of cement polymerization remain largely unknown. We hypothesized that this process is biochemically similar to blood clotting, a critical physiological response that is also based on aggregation and cross-linking of proteins. Like key elements of vertebrate and invertebrate blood clotting, barnacle cement polymerization was shown to involve proteolytic activation of enzymes and structural precursors, transglutaminase cross-linking and assembly of fibrous proteins. Proteolytic activation of structural proteins maximizes the potential for bonding interactions with other proteins and with the surface. Transglutaminase cross-linking reinforces cement integrity. Remarkably, epitopes and sequences homologous to bovine trypsin and human transglutaminase were identified in barnacle cement with tandem mass spectrometry and/or western blotting. Akin to blood clotting, the peptides generated during proteolytic activation functioned as signal molecules, linking a molecular level event (protein aggregation) to a behavioral response (barnacle larval settlement). Our results draw attention to a highly conserved protein polymerization mechanism and shed light on a long-standing biochemical puzzle. We suggest that barnacle cement polymerization is a specialized form of wound healing. The polymerization mechanism common between barnacle cement and blood may be a theme for many marine animal glues.

Shell Structure and Growth in the Base Plate of the Barnacle Amphibalanus amphitrite.

The base plate of the acorn barnacle Amphibalanus amphitrite (equivalent to Balanus amphitrite) is composed of hierarchically scaled, mutually aligned calcite grains, adhered to the substratum via layered cuticular tissue and protein. Acorn barnacles grow by expanding and lengthening their side plates, under which the cuticle is stretched, and adhesive proteins are secreted. In barnacles with mineralized base plates, such as A. amphitrite, a mineralization front follows behind, radially expanding the base plate at the periphery. In this study, we show that the new mineralization develops above the adhesion layers in a unique trilayered structure. Calcite crystallites in each of the layers have distinct sizes, varying from coarse-grained (>1 µm across) in the upper layer, to fine-grained (∼1 µm) in the middle layer, to nanoparticulate (∼40 nm) in the basal layer. The fine-grained crystallites dominate the growth front, comprising the bulk of the shell at the periphery, with later coarse grain development on the top of the base plate (toward the barnacle interior) and nanocrystalline calcite templating underneath in contact with the cuticle/protein layer. While the coarse-grained calcite on the upper surface contains a range of crystal orientations, the underlying fine-grained and nanocrystalline calcite are mutually oriented to within a few degrees of each other. Electron diffraction and X-ray absorption spectroscopy confirm that all of the crystallites are calcite, and metastable aragonite or amorphous calcium carbonate (ACC) phases are not observed. The complex morphology of the leading edge of the base plate suggests that crystallization initiates with the emplacement of mutually aligned fine-grained calcite, followed by the accumulation of coarser grains above and nucleation of highly oriented nanocrystalline grains below.

Identifying the dynamic compressive stiffness of a prospective biomimetic elastomer by an inverse method.

Soft elastomeric materials that mimic real soft human tissues are sought to provide realistic experimental devices to simulate the human body's response to blast loading to aid the development of more effective protective equipment. The dynamic mechanical behavior of these materials is often measured using a Kolsky bar because it can achieve both the high strain rates (>100s(-1)) and the large strains (>20%) that prevail in blast scenarios. Obtaining valid results is challenging, however, due to poor dynamic equilibrium, friction, and inertial effects. To avoid these difficulties, an inverse method was employed to determine the dynamic response of a soft, prospective biomimetic elastomer using Kolsky bar tests coupled with high-speed 3D digital image correlation. Individual tests were modeled using finite elements, and the dynamic stiffness of the elastomer was identified by matching the simulation results with test data using numerical optimization. Using this method, the average dynamic response was found to be nearly equivalent to the quasi-static response measured with stress-strain curves at compressive strains up to 60%, with an uncertainty of ±18%. Moreover, the behavior was consistent with the results in stress relaxation experiments and oscillatory tests although the latter were performed at lower strain levels.

Barnacles resist removal by crack trapping.

We study the mechanics of pull-off of a barnacle adhering to a thin elastic layer which is bonded to a rigid substrate. We address the case of barnacles having acorn shell geometry and hard, calcarious base plates. Pull-off is initiated by the propagation of an interface edge crack between the base plate and the layer. We compute the energy release rate of this crack as it grows along the interface using a finite element method. We also develop an approximate analytical model to interpret our numerical results and to give a closed-form expression for the energy release rate. Our result shows that the resistance of barnacles to interfacial failure arises from a crack-trapping mechanism.