11th International Conference on Forensic Inference and Statistics (ICFIS 2023)

1. Evidence, legal proof and the psychology of factfinding

The evaluation of evidence is a concept not restricted to forensic evidence evaluated against hypotheses on source or activity level. The ultimate evaluation of evidence is with the court and impacts along the chain of evidential reasoning down to the first discovery of a crime or identification of a legal issue.

2. Education, communication and the human factor

How do we (should we) educate or train forensic practitioners, police officers, prosecutors, defence attorneys, judges and jurors to enhance the understanding of forensic interpretation and evaluation and/or the application of statistical models to forensic data? What could be successful communication between forensic service providers and users, taking into account their widely differing backgrounds and roles.

3. Evaluating evidence

Computing evidential strength: probability, statistics, artificial intelligence and new technologies

What statistical models and techniques do we use and why? There is a trend that, per sample, the data become increasingly large. However, the number of samples often is still relatively small. Is artificial intelligence (machine learning) a way forward in more and more data-rich forensic casework? In areas that are sample-poor, what progress is made?

4. Impression evidence

Friction ridges, footwear marks, toolmarks, …

The development of LR-based models for fingerprints but also for other types of impression evidence has moved fast since the last conference in this series. What lessons can be learnt, and how do legal systems face this development, in particular when evidence is reported using likelihood ratios instead of categorically?

5. Inference beyond the source

Complete case evaluation, activity level, causality, …

What are the right questions to ask and what reasonable answers can we provide (and how)?

6. Interpreting evidence

Different models for evidence evaluation, their pros and cons and assessment

What types of models do we want (e.g. feature-based vs. score-based)? How do we assess how these models address the correct question and how do we define the “usefulness” of these models to support inference in the legal context?

7. Civil applications, statistical modelling and methods and data analysis outside the evidence evaluation framework

Forensic statistics as well as the more general Statistics and the law is not restricted to decision-making regarding guilt or innocence, alleged activities or the attribution of source or class. What are the challenges when it comes to sampling of forensic material, prediction of contents and purity of drug seizures, firing range estimation, distance measuring in images etc.?

Plenary talks will be given by

• Charles Berger, Netherlands Forensic Institute and Leiden University, The Netherlands
• Alicia Carriquiry, Director of CSAFE, Iowa State University, US
• Jeannette Leegwater, Netherlands Forensic Institute, The Netherlands
• Tacha Hicks Champod, University of Lausanne, Switzerland
• William Thompson, University of California Irvine, US

Scientific committee

• Alex Biedermann, Université de Lausanne, Switzerland
• Silvia Bozza, Università Ca’ Foscari, Venice, Italy
• JoAnn Buscaglig, FBI Laboratory, US
• Alicia Carriquiry, Iowa State University, US
• James Curran, University of Auckland, New Zealand
• Norman Fenton, Queen Mary University of London, UK
• Therese Graversen, University of Copenhagen, Denmark
• David Kaye, Pennsylvania State University, US
• David Lagnado, University College London, UK
• Anjali Mazumder, Alan Turing Institute, UK
• Petter Mostad, Chalmers University, Sweden
• Julia Mortera, Università Roma Tre, Italy
• Danica Ommen, Iowa State University, US
• Qing Pan, George Washington University, US
• Paul Roberts, University of Nottingham, UK
• Christopher Saunders, South Dakota State University, US
• Marjan Sjerps, University of Amsterdam/Netherlands Forensic Institute, Netherlands
• Hal Stern, University of California Irvine, US
• Franco Taroni, Université de Lausanne, Switzerland
• Peter Vergeer, Netherlands Forensic Institute, Netherlands
• Gerhard Wevers, Institute of Environmental Science and Research Limited (ESR), New Zealand
• Amy Wilson, University of Edinburgh, UK
• Grzegorz Zadora, Institute of Forensic Research and University of Silesia, Poland

Local organising committee

• Anders Nordgaard, Swedish Police Authority – National Forensic Centre, chair
• Christian Dahlman, Lund University, Faculty of Law, co-chair
• Christoffer Wong, Lund University, Faculty of Law
• Maria Aminoff, Lund University, Faculty of Law
• Birgitta Rådström, Swedish Police Authority – National Forensic Centre
• Jonas Malmborg, Swedish Police Authority – National Forensic Centre
• Ronny Hedell, Swedish Police Authority – National Forensic Centre

The following workshops will be organised (provided a sufficient number of registered attendees).

1. Footwear comparisons: Become an expert for a day

This workshop has been cancelled due to too few participants.

2. North Sea Group workshop on theory of legal evidence

Workshop on theoretical challenges at the intersection of law and forensic science. Example of questions to be discussed: What methodology should legal fact-finders use for assessing the combined effect of forensic results and other kinds of evidence? How can the legal standard of proof be understood in probabilistic terms? Presentations by invited speakers.

This workshop is organised parallel to ICFIS2023, but participants at ICFIS are welcome to join free of charge.

3. Opportunities with models for probabilistic genotyping

Recent statistical advances have allowed more complex DNA profiles to be used as evidence in court. While probabilistic genotyping comes with major challenges, adopting a statistical approach also creates the opportunity for presenting a much more detailed story about the DNA evidence. Many of the questions naturally posed about the evidence, are equally naturally addressed as part of a statistical analysis.

My DNAmixtures software (https://dnamixtures.r-forge.r-project.org/) has been used for statistical evaluation of mixed DNA profiles in criminal cases both in Denmark and in the United Kingdom. I will demonstrate how I have used statistical reasoning in witness statements to give a detailed picture of my interpretation of the DNA evidence and to argue the reliability of the stated LR.

I will use DNAmixtures for the demonstration and hands-on tutorials, but the statistical reasoning applies broadly to other models for probabilistic genotyping.

Participants should ideally bring their own laptops to facilitate practical exercises.

Knowledge of statistics and the R/RStudio software is an advantage, but not required.

• Organiser: Therese Graversen, IT University of Copenhagen
• Day and time frame: Wednesday 14 June 13:30 – 17:00
• Requirements: Attendees should bring their own laptops
• More information: Therese Graversen, theg [at] itu [dot] dk (theg[at]itu[dot]dk)

Note! If you have already registered for the conference and also wish to attend this workshop, you can send an e-mail to icfis [dot] registration [at] polisen [dot] se (icfis[dot]registration[at]polisen[dot]se). Write “ICFIS2023-Workshop 3” in the subject field and as message write “I wish to attend Workshop 3”.

There will be a number of special sessions with invited speakers/discussants.

There is also special call for papers on score-based approached (see bottom of this page).

Round-table discussion on “Healthcare serial killer or coincidence?”

Organiser

Julia Mortera, Professor of Statistics, Università Roma Tre and Honorary Professor University of Bristol

Confirmed speakers

• Peter Green, Emeritus Professor of Statistics & Professorial Research Fellow, University of Bristol
• Jane Hutton, Professor of Statistics, University of Warwick, UK
• William Thompson, Professor Emeritus of Criminology, Law, and Society, University of California, Irvine
• Lena Wahlberg, Associate Professor in Jurisprudence, Lund University

About

Suspicions about medical murder often arise due to a surprising or unexpected series of events, such as an unusual number of deaths among patients under the care of a particular professional. There have been several high-profile cases where healthcare professionals are accused of murdering patients where statistical evidence has been misused.

There are major concerns about the analysis and interpretation of the evidence in these types of investigations and whether it can be guaranteed that the data have been compiled in an objective and unbiased manner. When interpreting such data, investigators need to consider:

• Could the deaths have occurred for reasons other than murder?
• If murder was the cause, is the person under suspicion responsible?

Attention is rarely given to ensuring that unconscious bias has not influenced the selection of these cases. Experts informing an investigation should be kept “blind” to all aspects of the case. Blinding is a key tool in minimising prejudicial subjective effects such as unconscious bias.

There is a need for better collaboration between the legal and statistical communities to prevent miscarriages of justice happening in the future.

Statistical Models for Fingerprint Analysis: Thinking Broadly about the Future

Chair

Simon A. Cole, Professor, Department of Criminology, Law & Society, University of California Irvine, and Director National Registry of Exonerations

Confirmed speakers

• Alex Biedermann, Associate Professor, University of Lausanne
• Caroline Gibb, Researcher, University of Twente
• Geoffrey Stewart Morrison, Associate Professor of Forensic Speech Science, Aston University

About

Over the past several years, increasing attention has been drawn to the need for statistical models in forensic science. It has been argued that statistical models are necessary to assist forensic scientists by enabling them to evaluate and report the significance of their findings in a logical and scientifically defensible manner. However, the actual development and use of such models has been slow and subject to controversy. Forensic DNA profiling is perhaps the area in which such models have been most fully developed, but the ascendence of probabilistic genotyping has been characterized by vigorous controversy and debate.

Aside from DNA profiling, perhaps the most obvious candidate discipline for a statistical model is friction ridge (“fingerprint”) analysis. Fingerprint analysis remains widely used and highly trusted. Conceptual work on statistical models for fingerprint analysis has been done (e.g.,Neumann et al., 2012), and at least two working models are available (e.g.,Swofford et al., 2018).

The focus of this session is to envision what a future with a fingerprint statistical model will look like and to anticipate the scientific, legal, sociological, and ethical issues that such a future may entail.

Special open call: Applications and Foundational Methods in the Presentation and Interpretation of Evidence with Score-Based Approaches.

Organiser

Christopher P. Saunders, Professor of Statistics/NRM Affiliate Professor, south Dakota State University

About

In the broad field of forensic science, there are three main interpretation paradigms that examiners follow when evaluating evidence in the context of the identification of source problem. The first method is largely frequentist, often referred to as the Two-Stage or Classical approach and considers random matches; the second method uses a likelihood ratio/Bayes Factor focused on logical and coherent decisions, the final class of methods derive from the field of statistical pattern recognition as used in machine learning and AI, and broadly corresponds to Royall’s likelihoodist approaches. All three classes of methods rely on the use of probability distributions, whether the distributions characterize the belief about or the empirical/frequentist properties of the evidence-generating process.

However, in many settings, it is difficult to construct these probability distributions with respect to the natural feature space in which the evidence is observed, so it has become common practice to rely on a scoring function that represents the results of the comparison of two sets of the evidence as either a univariate similarity or dissimilarity score. Then, the induced relevant probability distributions are used for the probabilistic interpretation and presentation of evidence. Additionally, the score-based methods have an attractive advantage in that they allow the forensic scientist to determine how to discriminate (or characterize the similarity) between sources and the statistician to focus on the probabilistic statements induced by the interaction of the forensic propositions and score function.

For this program call, we are interested in presentations and panel discussions on both applications of score-based methods to forensic source ID problems that include a discussion of the performance of the approach (including calibration), applications of score-based approaches to specific modalities, development of scoring functions, and the theoretical aspects of score-based methods.

Welcome reception

Welcome reception on Monday 12 June (at conference venue) with drinks and canapés.

Poster session

Poster session on Tuesday 13 June (at conference venue) with drinks and snacks.

City tour and conference dinner

City tour and conference dinner on Wednesday 14 June.

• A guided walk in the city centre
• Conference dinner at AF-borgen (close to conference venue)

Registration is made using the registration form found below. All necessary information is given in that form.

The ordinary registration fee is SEK 6500 (incl. VAT) if registered before April 15, 2023.

There is a limited number of lower fees for students (SEK 3300 incl. VAT) provided a signed certification from supervisor is enclosed at registration (see registration form).

The fee includes:

• Admission to the conference
• Conference documents
• Morning and afternoon coffees/teas
• Lunches
• Welcome reception
• Poster session drinks

The payment of the fee will be by invoice.

Note! The registration fee is based on an estimated number of participants, which has been set fairly low. Should the number of registered participants be substantially higher than estimated there will be a reduction in the fee on the invoice.

If registration is made at latest on 15 April 2023, it will be possible for the organisers to have the payment confirmed before the conference (due to that the invoice process comprises regulated lead times). Registration after 15 April 2022 is possible (extended fee), but this requires that the invoice is paid immediately when it arrives despite the due date given.

The conference dinner fee is SEK 750 and is additional to the registration fee. This also includes a guided tour in the city of Lund.

Accompanying persons that wish to participate at the welcome reception and the conference dinner may be registered with a fee of SEK 900.

ICFIS 2023 Registration form (PDF, 208 kB, new tab)

Please, fill in and return the completed registration form by e-mail to: icfis [dot] registration [at] polisen [dot] se (icfis[dot]registration[at]polisen[dot]se). Write “ICFIS2023” in the subject field.

VAT rules and applicability

The Swedish VAT (Value Added Tax) rules for a physical conference imply that all delegates will be charged 25% VAT, but the VAT can be reclaimed in most cases (see info below). The VAT will be specified on the invoice. Hence, a valid VAT or tax registration number for your company organisation must be provided with the registration.

Delegates from outside Sweden

For participants from outside Sweden, the following rules apply:

• Within EU: A valid VAT number is mandatory, and it is important that your number is stated exactly according to EU standards. Further documentation must be provided on request. To validate your number before registering, please click here.
• Outside EU: A valid VAT number or equivalent tax registration number is mandatory. Any further documentation should be provided on request.

Delegates from Sweden

For Swedish delegates, representing a Swedish legal person or organisation, the VAT will be handled according to normal Swedish VAT rules.

People attending privately

Any person attending privately without representing a company/organisation are required to pay VAT (25%).

How to reclaim Swedish VAT

• Within EU: For delegates representing a company/organisation from an EU country, their company/organisation can reclaim the VAT through an EU electronic mechanism. For this reason, a valid VAT-number must be provided with the registration. The rules applicable for other EU countries can be studied in this Swedish Tax Agency link.
• Outside EU: For delegates from outside the EU, the possibility for their company/organisation to reclaim VAT also exists. Also, in this case, a valid VAT number or equivalent tax registration number is mandatory. The rules can be studied on the Swedish Tax Agency website on VAT refund and the form for VAT return can be downloaded here (PDF, new tab). Local regulations in the participant’s country of residence may also apply in addition to the Swedish VAT rules.