-
Directive Explanations for Actionable Explainability in Machine Learning Applications
Authors:
Ronal Singh,
Paul Dourish,
Piers Howe,
Tim Miller,
Liz Sonenberg,
Eduardo Velloso,
Frank Vetere
Abstract:
This paper investigates the prospects of using directive explanations to assist people in achieving recourse of machine learning decisions. Directive explanations list which specific actions an individual needs to take to achieve their desired outcome. If a machine learning model makes a decision that is detrimental to an individual (e.g. denying a loan application), then it needs to both explain…
▽ More
This paper investigates the prospects of using directive explanations to assist people in achieving recourse of machine learning decisions. Directive explanations list which specific actions an individual needs to take to achieve their desired outcome. If a machine learning model makes a decision that is detrimental to an individual (e.g. denying a loan application), then it needs to both explain why it made that decision and also explain how the individual could obtain their desired outcome (if possible). At present, this is often done using counterfactual explanations, but such explanations generally do not tell individuals how to act. We assert that counterfactual explanations can be improved by explicitly providing people with actions they could use to achieve their desired goal. This paper makes two contributions. First, we present the results of an online study investigating people's perception of directive explanations. Second, we propose a conceptual model to generate such explanations. Our online study showed a significant preference for directive explanations ($p<0.001$). However, the participants' preferred explanation type was affected by multiple factors, such as individual preferences, social factors, and the feasibility of the directives. Our findings highlight the need for a human-centred and context-specific approach for creating directive explanations.
△ Less
Submitted 2 February, 2021;
originally announced February 2021.
-
Distal Explanations for Model-free Explainable Reinforcement Learning
Authors:
Prashan Madumal,
Tim Miller,
Liz Sonenberg,
Frank Vetere
Abstract:
In this paper we introduce and evaluate a distal explanation model for model-free reinforcement learning agents that can generate explanations for `why' and `why not' questions. Our starting point is the observation that causal models can generate opportunity chains that take the form of `A enables B and B causes C'. Using insights from an analysis of 240 explanations generated in a human-agent ex…
▽ More
In this paper we introduce and evaluate a distal explanation model for model-free reinforcement learning agents that can generate explanations for `why' and `why not' questions. Our starting point is the observation that causal models can generate opportunity chains that take the form of `A enables B and B causes C'. Using insights from an analysis of 240 explanations generated in a human-agent experiment, we define a distal explanation model that can analyse counterfactuals and opportunity chains using decision trees and causal models. A recurrent neural network is employed to learn opportunity chains, and decision trees are used to improve the accuracy of task prediction and the generated counterfactuals. We computationally evaluate the model in 6 reinforcement learning benchmarks using different reinforcement learning algorithms. From a study with 90 human participants, we show that our distal explanation model results in improved outcomes over three scenarios compared with two baseline explanation models.
△ Less
Submitted 12 September, 2020; v1 submitted 28 January, 2020;
originally announced January 2020.
-
Explainable Reinforcement Learning Through a Causal Lens
Authors:
Prashan Madumal,
Tim Miller,
Liz Sonenberg,
Frank Vetere
Abstract:
Prevalent theories in cognitive science propose that humans understand and represent the knowledge of the world through causal relationships. In making sense of the world, we build causal models in our mind to encode cause-effect relations of events and use these to explain why new events happen. In this paper, we use causal models to derive causal explanations of behaviour of reinforcement learni…
▽ More
Prevalent theories in cognitive science propose that humans understand and represent the knowledge of the world through causal relationships. In making sense of the world, we build causal models in our mind to encode cause-effect relations of events and use these to explain why new events happen. In this paper, we use causal models to derive causal explanations of behaviour of reinforcement learning agents. We present an approach that learns a structural causal model during reinforcement learning and encodes causal relationships between variables of interest. This model is then used to generate explanations of behaviour based on counterfactual analysis of the causal model. We report on a study with 120 participants who observe agents playing a real-time strategy game (Starcraft II) and then receive explanations of the agents' behaviour. We investigated: 1) participants' understanding gained by explanations through task prediction; 2) explanation satisfaction and 3) trust. Our results show that causal model explanations perform better on these measures compared to two other baseline explanation models.
△ Less
Submitted 20 November, 2019; v1 submitted 26 May, 2019;
originally announced May 2019.
-
A Grounded Interaction Protocol for Explainable Artificial Intelligence
Authors:
Prashan Madumal,
Tim Miller,
Liz Sonenberg,
Frank Vetere
Abstract:
Explainable Artificial Intelligence (XAI) systems need to include an explanation model to communicate the internal decisions, behaviours and actions to the interacting humans. Successful explanation involves both cognitive and social processes. In this paper we focus on the challenge of meaningful interaction between an explainer and an explainee and investigate the structural aspects of an intera…
▽ More
Explainable Artificial Intelligence (XAI) systems need to include an explanation model to communicate the internal decisions, behaviours and actions to the interacting humans. Successful explanation involves both cognitive and social processes. In this paper we focus on the challenge of meaningful interaction between an explainer and an explainee and investigate the structural aspects of an interactive explanation to propose an interaction protocol. We follow a bottom-up approach to derive the model by analysing transcripts of different explanation dialogue types with 398 explanation dialogues. We use grounded theory to code and identify key components of an explanation dialogue. We formalize the model using the agent dialogue framework (ADF) as a new dialogue type and then evaluate it in a human-agent interaction study with 101 dialogues from 14 participants. Our results show that the proposed model can closely follow the explanation dialogues of human-agent conversations.
△ Less
Submitted 5 March, 2019;
originally announced March 2019.
-
Interaction Design for Explainable AI: Workshop Proceedings
Authors:
Prashan Madumal,
Ronal Singh,
Joshua Newn,
Frank Vetere
Abstract:
As artificial intelligence (AI) systems become increasingly complex and ubiquitous, these systems will be responsible for making decisions that directly affect individuals and society as a whole. Such decisions will need to be justified due to ethical concerns as well as trust, but achieving this has become difficult due to the `black-box' nature many AI models have adopted. Explainable AI (XAI) c…
▽ More
As artificial intelligence (AI) systems become increasingly complex and ubiquitous, these systems will be responsible for making decisions that directly affect individuals and society as a whole. Such decisions will need to be justified due to ethical concerns as well as trust, but achieving this has become difficult due to the `black-box' nature many AI models have adopted. Explainable AI (XAI) can potentially address this problem by explaining its actions, decisions and behaviours of the system to users. However, much research in XAI is done in a vacuum using only the researchers' intuition of what constitutes a `good' explanation while ignoring the interaction and the human aspect. This workshop invites researchers in the HCI community and related fields to have a discourse about human-centred approaches to XAI rooted in interaction and to shed light and spark discussion on interaction design challenges in XAI.
△ Less
Submitted 13 December, 2018;
originally announced December 2018.
-
An integrated P-T-H2O-lattice strain model to quantify the role of clinopyroxene fractionation on REE+Y/HFSE patterns of mafic alkaline magmas: Application to eruptions at Mt. Etna
Authors:
Silvio Mollo,
Jonathan Blundy,
Piergiorgio Scarlato,
Serena Pia De Cristofaro,
Vanni Tecchiato,
Flavio Di Stefano,
Francesco Vetere,
Francois Holtz,
Olivier Bachmann
Abstract:
A correct description and quantification of the geochemical behaviour of REE+Y (rare earth elements and Y) and HFSE (high field strength elements) is a key requirement for modelling petrological and volcanological aspects of magma dynamics. In this context, mafic alkaline magmas (MAM) are characterized by the ubiquitous stability of clinopyroxene from mantle depths to shallow crustal levels. On on…
▽ More
A correct description and quantification of the geochemical behaviour of REE+Y (rare earth elements and Y) and HFSE (high field strength elements) is a key requirement for modelling petrological and volcanological aspects of magma dynamics. In this context, mafic alkaline magmas (MAM) are characterized by the ubiquitous stability of clinopyroxene from mantle depths to shallow crustal levels. On one hand, clinopyroxene incorporates REE+Y/HFSE at concentration levels that are much higher than those measured for olivine, plagioclase, and magnetite. On the other hand, the composition of clinopyroxene is highly sensitive to variations in pressure, temperature, and melt-water content, according to exchange-equilibria between jadeite and melt, and between jadeite/Ca-Tschermak and diopside-hedenbergite. As a consequence, the dependence of the partition coefficient on the physicochemical state of the system results in a variety of DREE+Y/DHFSE values that are sensitive to the magmatic conditions at which clinopyroxenes nucleate and grow. In order to better explore magma dynamics using clinopyroxene, a new P-T-H2O-lattice strain model specific to MAM compositions has been developed. The model combines a set of refined clinopyroxene-based barometric, thermometric and hygrometric equations with thermodynamically-derived expressions for the three lattice strain parameters, i.e., the strain-free partition coefficient (D0), the site radius (r0), and the effective elastic modulus (E).
△ Less
Submitted 19 December, 2018;
originally announced December 2018.
-
Towards a Grounded Dialog Model for Explainable Artificial Intelligence
Authors:
Prashan Madumal,
Tim Miller,
Frank Vetere,
Liz Sonenberg
Abstract:
To generate trust with their users, Explainable Artificial Intelligence (XAI) systems need to include an explanation model that can communicate the internal decisions, behaviours and actions to the interacting humans. Successful explanation involves both cognitive and social processes. In this paper we focus on the challenge of meaningful interaction between an explainer and an explainee and inves…
▽ More
To generate trust with their users, Explainable Artificial Intelligence (XAI) systems need to include an explanation model that can communicate the internal decisions, behaviours and actions to the interacting humans. Successful explanation involves both cognitive and social processes. In this paper we focus on the challenge of meaningful interaction between an explainer and an explainee and investigate the structural aspects of an explanation in order to propose a human explanation dialog model. We follow a bottom-up approach to derive the model by analysing transcripts of 398 different explanation dialog types. We use grounded theory to code and identify key components of which an explanation dialog consists. We carry out further analysis to identify the relationships between components and sequences and cycles that occur in a dialog. We present a generalized state model obtained by the analysis and compare it with an existing conceptual dialog model of explanation.
△ Less
Submitted 20 June, 2018;
originally announced June 2018.
-
Diffusive exchange of trace elements between alkaline melts: implications for element fractionation and timescale estimations during magma mixing
Authors:
Diego González-García,
Maurizio Petrelli,
Harald Behrens,
Francesco Vetere,
Lennart A. Fischer,
Daniele Morgavi,
Diego Perugini
Abstract:
The diffusive exchange of 30 trace elements during the interaction of natural mafic and silicic alkaline melts was experimentally studied at conditions relevant to shallow magmatic systems. In detail, a set of 12 diffusion couple experiments have been performed between natural shoshonitic and rhyolitic melts from the Vulcano Island (Aeolian archipelago, Italy) at a temperature of 1200 °C, pressure…
▽ More
The diffusive exchange of 30 trace elements during the interaction of natural mafic and silicic alkaline melts was experimentally studied at conditions relevant to shallow magmatic systems. In detail, a set of 12 diffusion couple experiments have been performed between natural shoshonitic and rhyolitic melts from the Vulcano Island (Aeolian archipelago, Italy) at a temperature of 1200 °C, pressures from 50 to 500 MPa, and water contents ranging from nominally dry to ca. 2 wt. %. Concentration-distance profiles, measured by Laser Ablation ICP-MS, highlight different behaviours, and trace elements were divided into two groups: (1) elements with normal diffusion profiles (13 elements, mainly low field strength and transition elements), and (2) elements showing uphill diffusion (17 elements including Y, Zr, Nb, Pb and rare earth elements, except Eu). For the elements showing normal diffusion profiles, chemical diffusion coefficients were estimated using a concentration-dependent evaluation method, and values are given at four intermediate compositions (SiO2 equal to 58, 62, 66 and 70 wt. %, respectively). A general coupling of diffusion coefficients to silica diffusivity is observed, and variations in systematics are observed between mafic and silicic compositions. Results show that water plays a decisive role on diffusive rates in the studied conditions, producing an enhancement between 0.4 and 0.7 log units per 1 wt.% of added H2O. Particularly notable is the behaviour of the trivalent-only REEs (La to Nd and Gd to Lu), with strong uphill diffusion minima, diminishing from light to heavy REEs. Modelling of REE profiles by a modified effective binary diffusion model indicates that activity gradients induced by the SiO2 concentration contrast are responsible for their development, inducing a transient partitioning of REEs towards the shoshonitic melt.
△ Less
Submitted 5 June, 2018;
originally announced June 2018.
-
Glass stability (GS) of chemically complex (natural) sub-alkaline glasses
Authors:
G. Iezzi,
A. L. Elbrecht,
M. Davis,
F. Vetere,
V. Misiti,
S. Mollo,
A. Cavallo
Abstract:
Glass stability (GS) indicates the glass reluctance or ability to crystallise upon heating; it can be characterised by several methods and parameters and is frequently used to retrieve glass-forming ability (GFA) of corresponding liquids as the case with which such liquids can be made crystal free via melt-quenching. Here, GS has been determined for the first time on six sub-alkaline glasses havin…
▽ More
Glass stability (GS) indicates the glass reluctance or ability to crystallise upon heating; it can be characterised by several methods and parameters and is frequently used to retrieve glass-forming ability (GFA) of corresponding liquids as the case with which such liquids can be made crystal free via melt-quenching. Here, GS has been determined for the first time on six sub-alkaline glasses having complex (natural) compositions, the most widespread and abundant on Earth. KT, KH, KW, KLL and w2 GS parameters increase linearly and monotonically as a function of SiO2, with very high correlations. Moreover, Tx values and GS parameters highly correlate with GFA via Rc (critical cooling rate), previously determined with ex-situ cooling-induced experiments. Therefore, GS scales with GFA for natural silicate compositions. In addition, the in-situ Rc value of B100 measured with DSC results > 45 °C/min (> 2700 °C/h), broadly corroborating the Rc of about 150 °C/min (9000 °C/h) determined ex-situ. In turn, relevant solidification parameters on heating or cooling can be obtained by DSC investigations also for chemically complex (natural) systems, similar to simple silicate systems. These outcomes are relevant for lavas or magmas that re-heat glass-bearing volcanic rocks, as well as for fabricate glass-ceramic materials with desirable texture and composition of phases starting from abundant and very cheap raw volcanic rocks.
△ Less
Submitted 28 February, 2018;
originally announced February 2018.
-
Effects of CO2 flushing on crystal textures and compositions: experimental evidence from recent K trachybasalts erupted at Mt. Etna
Authors:
Marisa Giuffrida,
Francois Holtz,
Francesco Vetere,
Marco Viccaro
Abstract:
Changes in magmatic assemblages and crystal stability as a response of CO2 flushing in basaltic systems have been never directly addressed experimentally, making the role of CO2 in magma dynamics still controversial and object of scientific debate. We conducted a series of experiments to understand the response of magmas from Etna volcano to CO2 flushing. We performed a first experiment at 300 MPa…
▽ More
Changes in magmatic assemblages and crystal stability as a response of CO2 flushing in basaltic systems have been never directly addressed experimentally, making the role of CO2 in magma dynamics still controversial and object of scientific debate. We conducted a series of experiments to understand the response of magmas from Etna volcano to CO2 flushing. We performed a first experiment at 300 MPa to synthesize a starting material composed of crystals of some hundreds of m and melt pools. This material is representative of an initial magmatic assemblage composed of plagioclase, clinopyroxene and a water undersaturated melt. In a second step, the initial assemblage was equilibrated at 300 and 100 MPa with fluids having different XCO2fl . Our experiments demonstrate that flushing basaltic systems with fluids may drastically affect crystal textures and phase equilibria depending on the amount of H2O and CO2 in the fluid phase. Since texture and crystal proportions are among the most important parameters governing the rheology of magmas, fluid flushing will also influence magma ascent to the Earths surface. The experimental results open new perspectives to decipher the textural and compositional record of minerals observed in volcanic rocks from Mt. Etna, and at the same time offer the basis for interpreting the information preserved in minerals from other basaltic volcanoes erupting magmas enriched in CO2.
△ Less
Submitted 21 February, 2018;
originally announced February 2018.
-
Experimental constraints on the rheology, eruption and emplacement dynamics of analog lavas comparable to Mercury's northern volcanic plains
Authors:
Francesco Vetere,
Stefano Rossi,
Olivier Namur,
Daniele Morgavi,
Valeria Misiti,
Paolo Mancinelli,
Maurizio Petrelli,
Cristina Pauselli,
Diego Perugini
Abstract:
We present new viscosity measurements of a synthetic silicate system considered an analogue for the lava erupted on the surface of Mercury. In particular, we focus on the northern volcanic plains (NVP), which correspond to the largest lava flows on Mercury and possibly in the Solar System. High-temperature viscosity measurements were performed at both superliquidus (up to 1736 K) and subliquidus c…
▽ More
We present new viscosity measurements of a synthetic silicate system considered an analogue for the lava erupted on the surface of Mercury. In particular, we focus on the northern volcanic plains (NVP), which correspond to the largest lava flows on Mercury and possibly in the Solar System. High-temperature viscosity measurements were performed at both superliquidus (up to 1736 K) and subliquidus conditions (1569-1502 K) to constrain the viscosity variations as a function of crystallinity (from 0 to 28\%) and shear rate (from 0.1 to 5 s 1). Melt viscosity shows moderate variations (4-16 Pa s) in the temperature range of 1736-1600 K. Experiments performed below the liquidus temperature show an increase in viscosity as shear rate decreases from 5 to 0.1 s 1, resulting in a shear thinning behavior, with a decrease in viscosity of 1 log unit. The low viscosity of the studied composition may explain the ability of NVP lavas to cover long distances, on the order of hundreds of kilometers in a turbulent flow regime. Using our experimental data we estimate that lava flows with thickness of 1, 5, and 10 m are likely to have velocities of 4.8, 6.5, and 7.2 m/s, respectively, on a 5 degree ground slope. Numerical modeling incorporating both the heat loss of the lavas and its possible crystallization during emplacement allows us to infer that high effusion rates (>10,000 m3/s) are necessary to cover the large distances indicated by satellite data from the MErcury Surface, Space ENvironment, GEochemistry, and Ranging spacecraft.
△ Less
Submitted 16 August, 2017;
originally announced August 2017.
-
Exponential Decay Of Concentration Variance During Magma Mixing: Robustness Of A Volcanic Chronometer And Implications For The Homogenization Of Chemical Heterogeneities In Magmatic Systems
Authors:
Stefano Rossi,
Maurizio Petrelli,
Daniele Morgavi,
Diego González-García,
Lennart A. Fischer,
Francesco Vetere,
Diego Perugini
Abstract:
The mixing of magmas is a fundamental process in the Earth system causing extreme compositional variations in igneous rocks. This process can develop with different intensities both in space and time, making the interpretation of compositional patterns in igneous rocks a petrological challenge. As a time-dependent process, magma mixing has been suggested to preserve information about the time elap…
▽ More
The mixing of magmas is a fundamental process in the Earth system causing extreme compositional variations in igneous rocks. This process can develop with different intensities both in space and time, making the interpretation of compositional patterns in igneous rocks a petrological challenge. As a time-dependent process, magma mixing has been suggested to preserve information about the time elapsed between the injection of a new magma into sub-volcanic magma chambers and eruptions. This allowed the use of magma mixing as an additional volcanological tool to infer the mixing-to-eruption timescales. In spite of the potential of magma mixing processes to provide information about the timing of volcanic eruptions its statistical robustness is not yet established. This represents a prerequisite to apply reliably this conceptual model. Here, new chaotic magma mixing experiments were performed at different times using natural melts. The degree of reproducibility of experimental results was tested repeating one experiment at the same starting conditions and comparing the compositional variability. We further tested the robustness of the statistical analysis by randomly removing from the analysed dataset a progressively increasing number of samples. Results highlight the robustness of the method to derive empirical relationships linking the efficiency of chemical exchanges and mixing time. These empirical relationships remain valid by removing up to 80% of the analytical determinations. Experimental results were applied to constrain the homogenization time of chemical heterogeneities in natural magmatic system during mixing. The calculations show that, when the mixing dynamics generate millimetre thick filaments, homogenization timescales of the order of a few minutes are to be expected.
△ Less
Submitted 30 June, 2017;
originally announced June 2017.
-
Water-enhanced interdiffusion of major elements between natural shoshonite and high-K rhyolite melts
Authors:
Diego González-García,
Harald Behrens,
Maurizio Petrelli,
Francesco Vetere,
Daniele Morgavi,
Chao Zhang,
Diego Perugini
Abstract:
The interdiffusion of six major elements (Si, Ti, Fe, Mg, Ca, K) between natural shoshonite and a high-K calc-alkaline rhyolite (Vulcano island, Aeolian archipelago, Italy) has been experimentally measured by the diffusion couple technique at 1200°C, pressures from 50 to 500 MPa and water contents from 0.3 (nominally dry) to 2 wt%. The experiments were carried out in an internally heated pressure…
▽ More
The interdiffusion of six major elements (Si, Ti, Fe, Mg, Ca, K) between natural shoshonite and a high-K calc-alkaline rhyolite (Vulcano island, Aeolian archipelago, Italy) has been experimentally measured by the diffusion couple technique at 1200°C, pressures from 50 to 500 MPa and water contents from 0.3 (nominally dry) to 2 wt%. The experiments were carried out in an internally heated pressure vessel, and major element profiles were later acquired by electron probe microanalysis. The concentration-distance profiles are evaluated using a concentration-dependent diffusivity approach. Effective binary diffusion coefficients for four intermediate silica contents are obtained by the Sauer-Freise modified Boltzmann-Matano method. At the experimental temperature and pressures, the diffusivity of all studied elements notably increases with dissolved H2O content. Particularly, diffusion is up to 1.4 orders of magnitude faster in a melt containing 2 wt.% H2O than in nominally dry melts. This effect is slightly enhanced in the more mafic compositions. Uphill diffusion was observed for Al, while all other elements can be described by the concept of effective binary interdiffusion. Ti is the slowest diffusing element through all experimental conditions and compositions, followed by Si. Fe, Mg, Ca and K diffuse at similar rates but always more rapidly than Si and Ti. This trend suggests a strong coupling between melt components. Since effects of composition (including water content) are dominant, a pressure effect on diffusion cannot be clearly resolved in the experimental pressure range.
△ Less
Submitted 10 June, 2017;
originally announced June 2017.
-
Quantifying magma mixing with the Shannon entropy: application to simulations and experiments
Authors:
D. Perugini,
C. P. De Campos,
M. Petrelli,
D. Morgavi,
F. P. Vetere,
D. B. Dingwell
Abstract:
We introduce a new quantity to petrology, the Shannon entropy, as a tool for quantifying mixing as well as the rate of production of hybrid compositions in the mixing system. The Shannon entropy approach is applied to time series numerical simulations and high-temperature experiments performed with natural melts. We note that in both cases the Shannon entropy increases linearly during the initial…
▽ More
We introduce a new quantity to petrology, the Shannon entropy, as a tool for quantifying mixing as well as the rate of production of hybrid compositions in the mixing system. The Shannon entropy approach is applied to time series numerical simulations and high-temperature experiments performed with natural melts. We note that in both cases the Shannon entropy increases linearly during the initial stages of mixing and then saturates toward constant values. Furthermore, chemical elements with different mobilities display different rates of increase of the Shannon entropy. This indicates that the hybrid composition for the different elements is attained at different times generating a wide range of spatio-compositional domains which further increase the apparent complexity of the mixing process. Results from the application of the Shannon entropy analysis are compared with the concept of Relaxation of Concentration Variance (RCV), a measure recently introduced in petrology to quantify chemical exchanges during magma mixing. We derive a linear expression relating the change of concentration variance during mixing and the Shannon entropy. We show that the combined use of Shannon entropy and RCV provides the most complete information about the space and time complexity of magma mixing. As a consequence, detailed information about this fundamental petrogenetic and volcanic process can be gathered. In particular, the Shannon entropy can be used as complement to the RCV method to quantify the mobility of chemical elements in magma mixing systems, to obtain information about the rate of production of compositional heterogeneities, and to derive empirical relationships linking the rate of chemical exchanges between interacting magmas and mixing time.
△ Less
Submitted 14 September, 2016;
originally announced September 2016.
-
Non-linear deformation and break up of enclaves in a rhyolitic magma: a case study from Lipari Island (Southern Italy)
Authors:
Francois Holtz,
Sascha Lenne,
Guido Ventura,
Francesco Vetere,
Philipp Wolf
Abstract:
A dome from Lipari Island (Southern Italy) consists of 12 vol. percent of circular, elongated and folded latitic enclaves hosted in a rhyolitic matrix. The dm- to cm-scale enclaves are more deformed than the mm-scale blobs. The critical value of the ratio between the viscous forces, which allow deformation and eventually break up blobs of latitic magma, and the interfacial tension forces is larg…
▽ More
A dome from Lipari Island (Southern Italy) consists of 12 vol. percent of circular, elongated and folded latitic enclaves hosted in a rhyolitic matrix. The dm- to cm-scale enclaves are more deformed than the mm-scale blobs. The critical value of the ratio between the viscous forces, which allow deformation and eventually break up blobs of latitic magma, and the interfacial tension forces is larger than 0.29. The Reynolds number is < 5.3. The equivalent radius and the axial ratio of the enclaves follow power-law distributions. This feature suggests that the break up and stretching of magmas are non-linear, scale-invariant, probably cyclic processes. The coexistence of enclaves of different shape and the self-similar size distributions suggest that chaotic advection plays a major role in the formation of mingled magmas. Caution must be used when measuring the finite strain from enclave shapes because they may break apart during the deformation.
△ Less
Submitted 29 November, 2004;
originally announced November 2004.
