data modeling for socially responsible seafood
Océane Elia Boulais
Thesis Proposal for the degree of Master of Science in Media Arts and Sciences at the Massachusetts Institute of Technology Fall 2019
Thesis advisor: Joseph A. Paradiso
Reader: Neha Nerula
Reader: Kate Wing
Recent media revelations in the seafood sector that disclose cases of human rights violations behest novel ways to integrate and access social responsibility assessments throughout the supply chain for whole-chain visibility. This thesis proposes the technical design and early-stage prototype of TransparenSea, a data model for integrating and interfacing with social responsibility assessment data for multi-stakeholder supply chains in a sector that has spent decades working to improve environmental sustainability. The proof-of-concept will be co-developed with the NGO Conservation International and the social auditing non-profit, Verité. Data from pilot assessments will be integrating with existing auditing schemes currently used in the Eastern Pacific purse seine tuna fishery supply chain. For evaluation, the data model will be utilized to construct integration and analytics for a Fishery Improvement Program pilot in Ecuador to identify risk areas, visibility issues and information gaps that inhibit responsive interventions to human rights abuses in the supply chain.
The true system, the real system, is our present construction of systematic thought itself, rationality itself, and if a factory is torn down but the rationality which produced it is left standing, then that rationality will simply produce another factory. If a revolution destroys a systematic government, but the systematic patterns of thought that produced that government are left intact, then those patterns will repeat themselves in the succeeding government. There’s so much talk about the system. And so little understanding.
― Robert M Pirsig
Supply chains are the living networks that shape how we interact with the world. In its most basic form, a supply chain is the sequence of processes involved in the production and distribution of a commodity. Sustainable supply chain management is defined well by Carter and Rogers1 as
“the strategic, transparent integration and achievement of an organization’s social, environmental, and economic goals in the systemic coordination of key inter-organizational business processes for improving the long-term economic performance of the individual company and its supply chains,”
Although supply chains have existed for millennia, the past two decades have brought insight from the emerging body of knowledge on the practice and theory of Sustainable Supply Chain Management (SSCM)2. Seen in Fig.1, SSCM has brought us the “triple bottom line” framework, where sustainable supply chains sit at the intersection of three fields: economic sustainability, environmental sustainability, and social sustainability3. A critical strategic issue for sustainability in supply chains is the confirmation and verification that processes, products and activities within the supply chain meet specific sustainability criteria and certifications4. An approach to analyzing this scheme is the concept of whole-chain visibility, which is increasingly viewed as essential criteria for the long-term competitiveness of a supply network5.
The seafood sector is a particularly interesting use case for the whole-chain visibility approach because of its global relevance to food security and livelihoods; employing millions, the sector is the primary source of protein for 3 of 7 people globally 6. Furthermore, the global seafood sector has spent decades working to improve environmental sustainability, leaving behind efforts to integrate human rights and anti-slavery mechanisms until recent media revelations7 8. These disclosures have propelled the sector into a time of internal assessment and have created conditions to fuel impetus for stakeholders to develop solutions to reduce and eliminate social abuses. Non-governmental organizations (NGOs), such as Conservation International, have responded with collaboratively-designed social responsibility assessments that have cross-cutting impact for the private sector, NGOs and government organizations9. Fisheries and Aquaculture Improvement Projects (FIPs or AIPs) are some of the most vibrant catalysts for change in seafood. Major buyers rely on FIPs and AIPs to meet sustainability commitments for their sourcing, and the model is proliferating with rapid expansion and proven environmental impacts10. Using the Improvement Model as the vehicle to drive the importance of human rights protection is necessary to align private sector, nonprofit, and governmental actions to improve social performance.
In order to root social responsibility into the Improvement Project model, the sector must address a data interoperability and integration problem; whole-chain visibility is only possible once the results of the social assessments (at the harvesting and pre-processing stages) are merged with existing auditing data(at the processing and post distribution phases) in a secure and reliable manner. If integrated properly, the conglomeration of assessments can help identify risk areas, visibility issues and information gaps which would aid in the design of responsive interventions11. The rest of this thesis proposal will aim to discuss the supply chain, not just as a multi-agent system with explicit communication between entities, but also as a network of nodes that are increasingly complex and learn to adapt with the increasing interconnectedness of global supply networks12.
Under this pretense, the supply chain model is envisioned more kin to an organism that grows in multiple dimension rather than a vertical, rigid structure.
This thesis will describe the technical design of TransparenSea: a network protocol that employs a federated schema for database management and the interface for enhanced understanding of the connection between social audits in the seafood supply chain. The prototype will be co-developed with NGO Conservation International and non-profit Verité for integrating social data from pilot assessments with existing auditing schemes currently used in the supply chain, to evaluate the potential for whole-supply-chain visibility. The proof-of-concept will enable the interoperability of social data from assessments with existing benchmarking, traceability, inventory management, and auditing systems at the processing and post distribution stage, for full supply-chain integrated data and assurance systems.
What will be implemented (a federated database management system) and designed (a novel way of tracking eco-social progress for conservation efforts) are described in further detail below.
Federated database management system (FDBMS) for heterogenous database integration
At the core of this thesis is the need for interoperability between relational databases (piloted social assessment data; collected in the field in September) and existing other databases, namely object-based database (social audit data collected at the processing and post distribution phases).
Data sources live with each stakeholder in the supply chain, creating disparate silos, therefore propagating low visibility and information discontinuity. In a FIP, there is little incentive for the individual to centralize their data sources with other stakeholders simply because of the required resources such an endeavor would take. However, there is great interest on the client-side to see a consolidated database for analytics and processing. Such interest would come from the proactive consumer, business partners who source from the FIP along with the NGO and research community. A federated database system is a collection of cooperating yet autonomous and heterogeneous database systems13 and in theory is a good contender to manage a multi-stakeholder supply chain.
There are known architectural challenges one faces when using a federated DBMS as opposed to a traditional, centralized DBMS that must be considered. For instance, an FDBMS should not require significant changes to existing databases and applications, with the idea that the investment in existing databases should be guarded. However, this requirement brings potential conflicts in representation, application semantics, export correctness, minimality (absence of duplication) and understandability14. Known solutions for such conflicts lie in the art of schema coordination15. With this architectural framework in mind, the following schema depicts the prototype intention.
Interface: graph-based nodal visualization
As a network, the supply chain is modeled as a dynamic network; each node will carry FIP meta-data, notating socio-ecological progress through size and color as it is made available. Geographic mapping of the network will be a reach goal for this thesis work, as it would useful to visualize geospatial implications of social responsibility assessments.
Authentication through Electronic Monitoring
As an additional verification mechanism, I will model an on-vessel crew member recognition system using low-cost, and therefore low-resolution, cameras using computer vision and deep learning applications. The system will flag vessels with crew members who are not registered employees to identify potentially enslaved or exploited persons on the vessel while at sea for third party auditors to review once the boat is back in port. This computer vision-powered flagging mechanism could also be useful for conditions on a vessel that would be otherwise deemed unsafe. For instance, the lack of a buoy or readily available man-over board equipment would correlate to a potentially unsafe work environment.
Using OpenCV16 for the initially trained data set, I will then model the system on myself to verify that meta data tagging (e.g. name, age, crew member status, employee ID, etcetera) can be associated with the face detected.
To date, the proposed system will be implemented using SPARQL for querying, mySQL for data management, and react17 for iterating the network graph interface. Existing ecological FIP data will come from FishChoice (pending) and the SustainableFisheries-UW FIP database18. Existing and piloted social auditing data will be sourced from Conservation International and Verité.
Field Work and Landscape Assessments
As of November 2019, there have been three landscape assessments of traceability technology in the seafood sector done for this thesis work. For a comprehensive perspective on the current status of data movement and storage throughout the supply chain, the following field missions were completed.
Technological Innovation in Seafood
What was the beginning of this thesis work began as a feasibility review of blockchain technology in the seafood sector, homing in on flurry of pilots that have launched in the past four years. Currently a pending manuscript under peer-review, I’ve previewed the discussion done here and will figure out how to integrate it into this body of work in the near future.
Over the course of a week in Honolulu, the objective was to better understand the availability of potential data partners for integration, witness the operation of a fishery processing facility, note the movement and format of data tied to the tuna at the Pier 38 Auction and observe on-vessel electronic monitoring techniques. Heavy discussions around the co-development of this thesis work with Conservation International also occurred at the Hawai’i office.
In Ecuador, I supported the Conservation International social responsibility assessment workshop. Over the course of the week being, we were trained to effectively utilize the ‘Montery Framework’ social responsibility assessment tool, which is a collaboration effort between Conservation International, the Conservation Alliance for Seafood Solutions, and the Coalition for Socially Responsible Seafood. During our time in the training session, we generated an implementation plan for on-site deployment in Manta, Ecuador where we met with the crew members of an industrial purse seine vessel. Leveraging existing relations with the Ecuador-Conservation International team, we received collaborative feedback while developing the implementation strategy that was relevant to the community at hand.
Upon arrival in Manta, we had a week to interview the crew members and observe vessel conditions through an on-site visit. The interviews are currently under review and the results of the social assessments from the pilots that occurred at the harvesting and pre-processing stage will be integrated upon release from Conservation International and Verité.
The work completed for this thesis will be evaluated by FisheryProgress (pending), Conservation International and Verité. In an effort to properly scope this thesis work, the prototype will focus on the data integration and deployment for a single FIP. Deployment will be evaluated based on the ability to offer insights to the NGO-committee behind the FisheryProgress FIP profiles (below) during the development of a comprehensive social data reporting platform that enable the identification of key technologies to track and improve social performance in FIPs.
The implications of creating an interface where stakeholders are evaluated for their efforts (or lack thereof) to protect human rights within their supply chain is enormous. Questions that remain open as I continue exploring :
Are traditional scoring colors (red, yellow, green) and grades (A, B, C…) the correct mechanism for rating stakeholders in the Improvement Program model, where the assumption is that they are making progress over a long term span (five years)?
What are the relevant indicators that would focus on advising impactful action plans for social responsibility and avoid greenwashing19?
Existing open-source tooling to trace supply chain data dynamically from an environmental accountability perspective can be on found on platforms like TRASE, that focus heavily on geographic representation. Bext360 is a blockchain-based technology company focusing on supply chains such as coffee, seafood, timber, minerals, cotton and palm oil. Visual dashboard analytics are mentioned, but it is not open-sourced. To date, there is no know open-source tooling to visualize FIP-uploaded social audit data. The seafood slavery risk tool is operated by the Monterey Bay Aquarium Seafood Watch Program, Liberty Asia and Sustainable Fisheries Partnership .
By the end of November, I will have completed a mock up of the network graph and the pilot assessment data collected in Ecuador. By December I will be constructing the database management system laid out in the architecture slide we saw earlier to deploy it in January. Towards the beginning of February, I will evaluate the network graph through a user interface survey and run planned performance testing on the database management instance. Finally in March, I hope to have all of my data in order to begin writing the thesis until submission in May.
This work is supported by the Northrup Grumman-Conservation International Technology for Conservation (T4C) University Grant, the Elements Collective and the MIT Media Lab Consortium Fund