[Henoko Boat Capsizing Accident] Considering the "Questions" at the Bereaved Family Press Conference Part 4: Can AI Improve Questions at Bereaved Family Press Conferences
This series is not intended to criticize the individual reporters who asked the questions. Nor is it intended to speculate on the reporters' intentions.
However, not blaming individuals is different from obscuring the problems with the questions. Based on the published Q&A, I will consider whether the questions were in line with the purpose of the press conference, whether they revealed new facts, and whether they were worth the burden placed on the bereaved family.
It is precisely because the parents answered sincerely that we were able to learn so much. However, the fact that they answered does not mean that it was right to ask those questions.
What became clear by organizing the 17 Q&As
In Part 0, I organized the Q&A with 17 questioners regarding the press conference held by Tomoka Takeishi's parents.
There were questions that were in line with the purpose of the press conference and revealed important facts.
On the other hand, there were also questions that delved into the parents' private grief rather than the cause of the accident or where responsibility lay. There were also instances where reporters asked again about things that had already been explained in the bereaved family's note or in their answers during the press conference. There were also instances where a single reporter asked several consecutive questions that were heavy both mentally and legally.
Listening to the press conference, one can feel a sense of discomfort with each question.
However, looking back at the entire press conference, which lasted over an hour,
who asked what
which questions were repetitive
what the bereaved family had already shared in their note
what was newly revealed by the questions
whether there were any questions that strayed from the purpose of the press conference
it takes a considerable amount of time to organize these things manually.
Could AI be used for such tasks?
I used AI for this organization myself
In creating this series, I had AI read the Q&A from the press conference and organized the "summary of the question," "summary of the answer," and "what was confirmed" for each questioner.
As a result, I was able to get an overview of the long press conference.
Questions that delved into the cause of the accident and safety management.
Questions that clarified the legal significance of the criminal complaint.
Questions that overlapped with content already discussed on note.
Questions that intruded into the parents' private lives and grief.
Questions that layered multiple points of discussion into a single statement.
By lining these up, it became clear where the sense of discomfort I felt while listening to the press conference originated.
However, it was not the AI that first identified the problem.
It was I who felt, while listening to the press conference, 'Is this really something that should be asked of the parents?' The AI played the role of comparing that discomfort against the entire Q&A session and organizing it into words.
A person feels discomfort, the AI organizes the whole, and a person makes a judgment once again.
This process was a back-and-forth of that nature.
What can be done before the press conference
I believe AI is most useful before a press conference begins.
If you have AI organize publicly available materials—such as the bereaved family's note, accident investigation reports, documents released by schools and government agencies, and past news reports—you can create a list like the following in a short amount of time.
Facts that have already been confirmed
Things the bereaved family has already explained
Points where content differs depending on the material
Things that have not yet been clarified
Things that should be newly confirmed through the criminal complaint
Things that should be asked of parties other than the bereaved family
Questions considered to have a high psychological burden
If you think about your questions based on this, you can ask about what lies ahead rather than having them explain content that has already been made public from the beginning.
For example,
What kind of daughter was Chika?
Instead of asking,
In your notes, you repeatedly wrote about how Chika cherished her school and teachers. How did that lead to the conflict you felt before deciding to sue the teachers?
you could ask this instead.
By organizing publicly available materials, you can not only reduce the number of questions but also transform them into deeper inquiries.
Reviewing questions in advance
You can also have AI read through the draft questions prepared by reporters to review them before they are asked.
The items to check are not that complex.
Does it align with the purpose of the press conference?
Can it reveal new facts or perspectives?
Has it already been answered in public documents?
Is it necessary to ask the person directly?
Does it intrude excessively into their private life?
Is there a possibility of unnecessarily forcing them to relive their grief?
Does a single statement contain multiple questions?
Can it be rephrased in a way that is less burdensome?
If you have AI check the questions asked at this press conference from these perspectives, you can organize them as follows, for example.
What could the school and the council have done to prevent the accident?
This question is highly relevant to the purpose of the press conference and reveals new facts or the thoughts of the bereaved family. It leads to identifying the cause of the accident and preventing recurrence, and it does not force the bereaved family to talk about their private lives unnecessarily.
What would you like to say to your daughter who passed away?
While the psychological burden is extremely high, there is almost no new information revealed regarding the cause of the accident, criminal liability, or the prevention of recurrence. It is necessary to re-examine whether the public interest gained through such questions is balanced against the burden placed on the bereaved family.
How do you spend the monthly anniversary of the death?
This question intrudes significantly into the family's private life and has a weak connection to the purpose of the press conference regarding the criminal complaint. Unless the bereaved family chooses to speak about it themselves, the necessity of asking such questions in a public forum must be carefully judged.
Why has there been no reply to the council for about three weeks?
There is some merit in confirming whether the exchange is ongoing. However, there is little public interest in pressing the bereaved family for the reason behind a delayed response.
There is no need to adopt the evaluations provided by AI exactly as they are.
Even so, they can serve as material to help one pause and consider, 'Is it really necessary to ask this?' before posing a question.
Redundancy during the press conference can also be checked
AI can also be utilized during the press conference itself.
By transcribing audio in real-time and organizing the questions and answers that have already been given, one can identify:
That question has already been answered
The same content is written in the bereaved family's note
One can narrow down the focus to only the parts that need new confirmation
A single statement contains three questions
It is a follow-up question on a topic the bereaved family has declined to answer
These things can be confirmed on the spot.
The moderator can also say:
That point has already been answered. Please narrow your focus to the unanswered parts.
Please limit your question to one.
As this concerns the private lives of your family, answering is optional.
It becomes easier to organize in this way.
In this instance, after the bereaved family declined to answer regarding how the footage was obtained, a follow-up question was asked that seemed to speculate on the source. A lawyer intervened, explaining that it should be treated as 'provided by the bereaved family'.
To what extent should follow-up questions be asked regarding content for which an intention to decline answering has been shown? If there is an AI-generated record, one can make a judgment after accurately confirming the immediately preceding answer.
Can also be used to support the 'one person, one question' rule
At this press conference, there were instances where a single reporter would throw out multiple questions at once.
I believe that on the side of those asking questions, there are circumstances where they want to confirm as much as possible during a limited opportunity.
However, the people answering are parents who have only lost their daughter four months ago. While tracing painful memories, they must organize and answer multiple questions while also considering legal implications and the impact on the school and students.
If you have AI check the question text,
This statement contains four items: defamation, civil litigation, negotiations with the council, and an apology.
It can be broken down like this.
Furthermore, one can select the single question most relevant to the purpose of the press conference and save the rest for another opportunity.
The 'one person, one question' rule is not a restriction intended to hinder reporters' coverage.
It is a consideration to ensure that the bereaved family can answer a single question calmly. And it leads to securing opportunities for other reporters to ask questions as well.
AI cannot replace empathy
Having written this far, it might seem as though it would be fine to let AI judge the appropriateness of questions.
However, we cannot leave judgment solely to AI.
AI can find repetitions in text, categorize questions, and point out the possibility of psychological burden.
On the other hand, it cannot feel the atmosphere of the press conference venue as it is.
The trembling of the parents' voices, the time they take to search for words, the confusion at the moment they receive a question, and the way the couple sitting next to each other supports one another.
It is the people present at the press conference who must decide, upon seeing such things, that 'it is better not to ask any further'.
Even if AI determines that a question is 'permissible,' if the person in front of you is suffering, you must stop asking.
AI is not a substitute for empathy.
Rather, I believe it should be used to create space for reporters to look at the person in front of them by leaving document organization and duplicate checking to AI.
AI also makes mistakes
Caution is also necessary when using AI.
AI may infer things not written in documents, combine different statements into one, or mistake who said what.
Therefore,
Always cross-check AI summaries with the original source material
Verify quotes of statements against the original text
Separate facts from inferences
Do not easily input non-public personal information
Final questions must be decided by reporters and editors
These procedures are necessary.
Information regarding victims and bereaved families in particular must be handled with care. It would be meaningless if using AI resulted in a different kind of privacy violation.
AI is not a device for determining the correct answer, but is appropriately used as an assistant to reduce oversights and redundancies.
Not for softening tough questions
Regarding the use of AI to check questions, I think there is also a concern that 'it might make reporters' questions too safe'.
However, there is no need to soften questions that are necessary.
Why did the school not conduct a preliminary inspection?
Why was there no safety meeting held?
Why did the supervising teacher not board the boat?
Why were no improvements made even though students had felt danger during past boat trips?
What was explained to the bereaved families and students after the accident?
Such questions should be directed strictly at the school, the school corporation, the council, and the administration.
AI-based inspection is not intended to weaken the pursuit of responsibility.
It is intended to make it clearer who should be asked what.
Do not direct harsh questions that should be aimed at organizations and those in charge toward the bereaved families.
Instead of spending time drawing out sorrow from the bereaved families, direct the power of reporting toward uncovering the structures that created the accident.
For that purpose as well, there is value in using AI for organizing materials and selecting questions.
Pause before asking a question
AI will not solve every problem.
Many of the questions felt problematic in this press conference could have been thought through if one had just paused for a moment, even without AI.
Read the bereaved family's note.
Confirm the purpose of the press conference.
Listen carefully to previous answers.
Do not ask multiple questions at once.
Do not pursue further if they have refrained from answering.
Ask about the causes and responsibilities of the accident rather than making them talk about their grief.
All of these are things that people can do in the first place.
Even so, oversights and repetitions occur amidst the limited time, vast amounts of material, and the tension surrounding question slots.
That is precisely why AI can be used as a 'second pair of eyes' to pause and reflect before asking a question.
Has this question already been answered?
Is this really something that should be asked of the bereaved family?
What new information will be revealed?
Is it worth the burden placed on the other party?
AI presents these questions, and ultimately, a human makes the judgment.
With that combination, wouldn't it be possible to achieve both the quality of reporting and consideration for the bereaved family?
In conclusion
The parents did not stand at the press conference to show their grief.
They did so because they believed there were things that needed to be conveyed to society in order to clarify the responsibility for why Chika-san had to lose her life and to ensure that the same accident never happens again.
To truly accept those feelings, we need to look back not only at how the parents answered, but also at what we asked.
A reporter's question can sometimes reveal the structure of an accident.
On the other hand, a question can sometimes force the bereaved family to expose grief that they did not need to share in a public setting.
Thinking about that boundary is not just a task for the reporters covering the story. We, as the recipients of the news, also have a responsibility to consider 'why this scene was reported' and 'how these words were elicited.'
AI will not decide the answer to that.
However, by organizing vast amounts of information, finding repetitions, and re-examining the necessity of questions, it can create opportunities for people to pause and reflect.
And in the end, it is people who make the judgment.
Thinking about what must be asked for the sake of society, while still imagining the pain of the person in front of you.
I believe the meaning of using AI is not to have it ask questions in place of humans, but to create the breathing room for humans to ask better questions.
The fact that the parents answered sincerely does not mean that it was right to ask those questions.
Without forgetting that, I hope that when the next similar press conference is held, it will be a space where both the questioner and the respondent are treated with a little more care than this time.
From Part 0, I have been thinking about the 'questions' exchanged at the bereaved family press conference. This concludes this series. Thank you for reading until the end.
*This article was written based on the Q&A from the press conference published by the Okinawa Times and the note published by the bereaved family. Please check the original articles for the full text of the Q&A and the statements.