The Oikonomia Network has published a new eBook, Faithfully Rendering God’s Word: Artificial Intelligence, Cultural Bias and the Future of Faithful Bible Translation. The book features essays by five theological scholars from our network, as well as reflections from leaders in Bible translation firms. The contributors are thinking together about how the use of Artificial Intelligence (AI) tools in Bible translation changes the landscape for some of the perennial challenges involved in translation work – most importantly cultural contextualization and bias, the danger of potentially colonizing power relationships, and the question of orthodoxy.

The ON has co-published the eBook with Seed Company, an innovative accelerator for Bible translation within the Wycliff family of organizations. Last fall, we shared the news that through our Karam Labs initiative, Seed Company had hired the Oikonomia Network to gather scholars from our network with their translation professionals for collaborative thought about these challenges. On February 19-20, about thirty people gathered for a two-day consultation event at the Seed Company headquarters in Dallas. The new eBook is the fruit of that collaborative time.

Zipping Up Our Knowledge

The book begins with a foreword by Seed Company CEO Samuel Chiang, and an introduction by ON Director Greg Forster. Both address the need for scholars and Bible translators to reconnect the theological knowledge stewarded in the academy and the work of Bible translation in the field. As Chiang puts it, AI is bringing about “a major shift in epistemology” in our culture that the church, in Bible translation and everywhere else, must consider carefully.

Making AI tools available to human translators to assist their work has enormous potential power to accelerate translation, but it raises urgent questions about issues like cultural bias and orthodoxy. In our newsletter story last fall we shared more details about how Seed Company is starting to use AI tools, and how they can be beneficial but also challenging.

Chiang uses the metaphor of a zipper, which meshes two separate sets of metal teeth together to form a bond. The theological academy and the work of Bible translation, Chiang observes, have become “unzipped” from each other, to the detriment of both. Our Karam Labs event with Seed Company was an opportunity to start zipping up the church’s knowledge bases.

Forster writes about the theological basis of this collaboration. The unzipping of academic theology and Bible translation is only one example of the way in which theory and practice tend to become detached in the advanced modern world. “Theological knowledge,” he writes, “has a unique power to counteract this fragmentation…It reconstructs connections between our daily practices and the big picture, reconnecting the fragments of life into a coherent whole.”

Four Theological Essays

The heart of the book is four theological essays written by ON scholars. These essays canvass the challenge of AI at the levels of theory and practice.

Larisa Levicheva and Abson Joseph of Wesley Seminary provide an overview of the cultural contextualization and bias challenges involved in Bible translation, and consider how the use of AI tools might be both helpful and challenging in this area. “Bible translation is the work of the church,” they remind us, and the Lord has equipped the church to carry out that work through cross-cultural collaboration. Levicheva and Joseph identify possible ways in which AI tools can help facilitate better collaboration in translation work across cultures, while also highlighting dangers associated with cultural bias, and processes that can reproduce colonial power dynamics. Silicon Valley leaders have already identified the potential of AI tools to incorporate the cultural biases of their programmers as a key concern. On the other hand, if accompanied by equal access and proper training, AI tools could make it possible to give Christians in the majority world more power and control over their own translation processes than is feasible under older models.

Hélene Dallaire of Denver Seminary provides a deep dive into the interdependence of culture and language, through the lens of how emotions are experienced and expressed differently in different languages and cultures. Emotions are embodied in human beings, who are cultural creatures; by comparison, even the best dictionary has a very limited and artificial ability to suggest the real emotional resonance of various terms in the biblical languages. Moreover, Dallaire provocatively asserts that “the most important element of Bible translation is the target language and not the source text.” Great sensitivity to the cultural context of the language into which the Bible is being translated must be a high priority. Dallaire reviews the emerging discourse on AI and emotions, concluding that while the ability of AI tools to understand emotions is currently very limited (she confirms this empirically through a series of amusing dialogues with an Alexa device), it is reasonable to expect that progress will be made on this front – no doubt faster than we will be comfortable with. The time for thinking carefully about these developments is now.

Mark Strauss of Bethel Seminary considers what practical processes would be needed to use AI tools responsibly. His central recommendation is an iterated feedback process. Translators could use AI tools to translate Bible passages into a language with a long-established translation tradition, such as English. The AI-assisted translation could then be compared to standard translations, with lessons drawn for how to improve the AI-assisted process – and, just as possibly, new lessons for traditional translators as well! The AI-assisted process could then be repeated, with the adjustments suggested by the original comparison, for the same passages into the same language. This would tell us whether and how the changes to the process improved the results. “If there are indeed rules and patterns behind these [culturally embedded] language choices,” writes Strauss, “then the translations should gradually improve even when the AI program has to render expressions it has never encountered before.”

Craig Blomberg of Denver Seminary completes the journey from theory to practice by reviewing a series of specific translation challenges, highlighting the book of Matthew but including other passages. As he ironically asks in the title of his paper, “How Hard Can It Be to Translate the Bible?” Pretty hard, it turns out; Blomberg highlights a wide variety of issues that might trip up an AI tool. These challenges include not only such obvious cultural factors as emotions and idiomatic expressions, but also less obvious but equally urgent issues, such as the ways in which translation work is interdependent with theological and ethical discourse in the church. For example, Blomberg challenges translators to consider their ethical responsibilities to Bible readers in the ways they might choose to render Matthew 5:39, in light of the fact that misunderstanding of this passage has been a contributing factor in domestic violence. He then asks how relying on AI tools might tend to affect translators’ sensitivity to that kind of question. Blomberg also lists a number of specific, tangible ways in which Bible translation companies can support indigenous translation workers well or poorly – translation firms’ processes of recruitment, training and working conditions can empower indigenous translators, or establish colonial power relationships.

Takeaways and Next Questions

In the eBook’s conclusion, Forster summarizes the most prominent themes that rose to the surface during the February consultation event:

  • Careful and ethical human stewardship of AI is essential, both on the “front end” where AI tools are created and on the “back end” where they are used. It is profoundly dangerous to think of AI as an autonomous decision-maker beyond human control. “In the Bible,” Forster writes, “idolatry is described as a practice of attributing independent agency to inert objects made of wood, metal and stone; attributing independent agency to objects made of sand would be just another instance.”
  • None of the challenges rising to the surface in the AI discussion is really new. On the contrary, every one of them – cultural context, power dynamics, concern for orthodoxy, etc. – is an age-old problem in Bible translation. AI does change the landscape, and raises these old problems in new (sometimes very new) forms, but the underlying challenges are the same ones the church has wrestled with for centuries. Therefore, the church has established wisdom to draw upon when considering these challenges. At the same time, we must be proactive about ensuring that all Christians (not just the traditional “old guard”) are included in the church’s collaboration.
  • The challenge of AI is a general challenge for our whole culture, not just a specific challenge for Bible translation. AI tools are already reshaping almost every structure in our cultural life. How can Christians be thoughtful about the way these changes are forming them – and how can Christians contribute proactively to the formation of these changes? Bible translation is a great place to start the conversation about responsible use of AI, but there is no need for the conversation to end there.

A list of “takeaways” and “next questions” that were gathered from participants in the concluding session of the event is also included in the eBook. An afterward by Shawn Ring of the AI firm Avodah points forward to the emerging conversation about these issues.

The ON is proud to have helped facilitate this important conversation! Check out Faithfully Rendering God’s Word: Artificial Intelligence, Cultural Bias and the Future of Faithful Bible Translation.